(b) Applying part (a) to the family {Af : i E /},we obtain that
Taking the complement of each side yields the second law.
(Ui
Aft =
ni(Af)c
ni
Ai.
2. Clearly
(i) An B = (Ac U Bc)c,
(ii) A\ B =An Be= (Ac U B)c,
(iii) A D. B =(A\ B) U (B \A)= (Ac U B)c U (AU Bc)c.
Now :F is closed under the operations of countable unions and complements, and therefore each of
these sets lies in :F.
3. Let us number the players 1, 2, ... , 2n in the order in which they appear in the initial table of draws.
The set of victors in the first round is a point in the space Vn = { 1, 2} x {3, 4} x .. x {2n  1, 2n}.
Renumbering these victors in the same way as done for the initial draw, the set of secondround
victors can be thought of as a point in the space VnI, and so on. The sample space of all possible
outcomes of the tournament may therefore be taken to be Vn x Vn I x x VI, a set containing
.
22n122n2 . . . 2I = 22nI pomts.
Should we be interested in the ultimate winner only, we may take as sample space the set
{1, 2, ... , 2n} of all possible winners.
4. We must check that g, satisfies the definition of a afield:
(a) 0 E :F., and therefore 0 = 0 n B E g,,
(b) if AI, A2, ... E :F., then Ui(Ai n B)= (Ui Ai) n BEg,,
(c) if A E :F., then Ac E :Fso that B \(An B)= Ac n B E g,.
(a), (b), and (d) are identically true; (c) is true if and only if A~ C.
(i) We have (using the fact that lP' is a nondecreasing set function) that
lP'(A
n B) = lP'(A)
+ lP'(B) 
135
+ lP'(B) 
1=
fz.
[1.3.2][1.3.4]
Solutions
Also, since An B ~A and An B ~ B, li"(A n B) :::: min{lP'(A), lP'(B)} = ~These bounds are attained in the following example. Pick a number at random from {1, 2, ... , 12}.
Taking A = {1, 2, ... , 9} and B = {9, 10, 11, 12}, we find that An B = {9}, and so li"(A) =
i.
li"(B) = ~. li"(A n B)
B = {1, 2, 3, 4}.
= /2 .
(ii) Likewise we have in this case lP'(A U B) :::: min{lP'(A) + li"(B), 1} = 1, and lP'(A U B) 2::
max{lP'(A), lP'(B)} =
These bounds are attained in the examples above.
2.
= n+oo
lim lP'(no head in first n tosses) = lim 2n = 0,
n+oo
n+oo
= 1
1
n+oo
lim (1 
n+oo
rk( =
1.
3. Lay out the saucers in order, say as RRWWSS. The cups may be arranged in 6! ways, but since
each pair of a given colour may be switched without changing the appearance, there are 6!: (2 !) 3 = 90
distinct arrangements. By assumption these are equally likely. In how many such arrangements is
no cup on a saucer of the same colour? The only acceptable arrangements in which cups of the
same colour are paired off are WWSSRR and SSRRWW; by inspection, there are a further eight
arrangements in which the first pair of cups is either SW or WS, the second pair is either RS or SR,
and the third either RW or WR. Hence the required probability is 10/90 = ~.
n Am+1}
Am+1) }
Using the induction hypothesis, we may expand the two relevant terms on the righthand side to obtain
the result.
136
Solutions
Conditional probability
[1.3.5][1.4.1]
Let A 1 , A 2, and A 3 be the respective events that you fail to obtain the ultimate, penultimate, and
antepenultimate ViceChancellors. Then the required probability is, by symmetry,
3
1 11"(
= 1 3(~)6 + 3(~)65.
(~)6.
= 1 lim
n+oo
6.
7.
1.
r=1
n As)
= np
~n(n 1)q.
Hence
r<s
L
li"(Ar
r<s
t<u
(r,s)#(t,u)
n As nAt n Au)+
By a careful consideration of the first three terms in the latter series, we find that
Hence
i = np G)x, so that p 2:: 3/(2n). Also, (~)q = 2np i. whence q .:S 4/n.
1.4 Solutions. Conditional probability
1.
I B) _ li"(A n B) _ li"(B n A)

li"(B)
li"(A)
137
li"(A) _ li"(B
li"(B) 
I A) li"(A)
li"(B)
[1.4.2][1.4.5]
Solutions
I B)
lP'(B
I A)
lP'(B)
lP'(A)
Set Ao = Q for notational convenience. Expand each term on the righthand side to obtain
3.
Let M be the event that the first coin is doubleheaded, R the event that it is doubletailed, and
Let H{ be the event that the lower face is a head on the ith toss, TJ the
event that the upper face is a tail on the ith toss, and so on. Then, using conditional probability ad
nauseam, we find:
(i)
lP'(Hl) = ~lP'(Hll M)
I
lP'(H/ n H.})
lP'(M)
2/3
2
I
= I = 5 5 = 3
lP'(Hu)
lP'(HI )
(ii)
lP'(HI I Hu) =
(iii)
+ ilP'(N
I HJ)
I + 2I( 1  lP' (II
I) = 32
= lP' ( HIII Hu)
HI Hu)
+ 2I . 3I = o5
(iv)
~
4
= 2I = .
4 . lP'(N)
5 + TO
5
lP'(M)
1 . lP'(M)
(v) From (iv), the probability that he discards a doubleheaded coin is ~, the probability that he
discards a normal coin is ~ (There is of course no chance of it being doubletailed.) Hence, by
conditioning on the discard,
lP'(H~) = ~lP'(H~ I M)
+ !lP'(H~
IN)= ~Ct
+ i. i) +Hi+ i.
i) = M
4. The final calculation of ~ refers not to a single draw of one ball from an urn containing three, but
rather to a composite experiment comprising more than one stage (in this case, two stages). While it
is true that {two black, one white} is the only fixed collection of balls for which a random choice is
black with probability ~, the composition of the urn is not determined prior to the final draw.
After all, if Carroll's argument were correct then it would apply also in the situation when the urn
originally contains just one ball, either black or white. The final probability is now
implying that
the original ball was one half black and one half white! Carroll was himself aware of the fallacy in
this argument.
i,
5. (a) One cannot compute probabilities without knowing the rules governing the conditional probabilities. If the first door chosen conceals a goat, then the presenter has no choice in the door to be
opened, since exactly one of the remaining doors conceals a goat. If the first door conceals the car, then
a choice is necessary, and this is governed by the protocol of the presenter. Consider two 'extremal'
protocols for this latter situation.
(i) The presenter opens a door chosen at random from the two available.
(ii) There is some ordering of the doors (left to right, perhaps) and the presenter opens the earlier
door in this ordering which conceals a goat.
Analysis of the two situations yields p = ~ under (i), and p =
138
i under (ii).
Solutions [1.4.6][1.5.3]
Independence
1, ],
Let a E [ ~ and suppose the presenter possesses a coin which falls with heads upwards with
probability f3 = 6a  3. He flips the coin before the show, and adopts strategy (i) if and only if the
coin shows heads. The probability in question is now ~ f3 + 1  /3) = a.
You never lose by swapping, but whether you gain depends on the presenter's protocol.
(b) Let D denote the first door chosen, and consider the following protocols:
(iii) If D conceals a goat, open it. Otherwise open one of the other two doors at random. In this
case p = 0.
(iv) If D conceals the car, open it. Otherwise open the unique remaining door which conceals a
goat. In this case p = 1.
As in part (a), a randomized algorithm provides the protocol necessary for the last part.
1(
6.
7. Let Ci be the colour of the ith ball picked, and use the obvious notation.
(a) Since each urn contains the same number n  1 of balls, the second ball picked is equally likely to
be any of the n (n  1) available. One half of these balls are magenta, whence 11"( C2 = M) =
(b) By conditioning on the choice of urn,
lP'(C2 = M I C 1 = M) = lP'(CJ, C2 = M) =
(n r)(n r  1)
lP'(CJ = M)
r=l n(n l)(n 2)
j!
Clearly
li"(Ac n B)= li"(B \{An B}) = lP'(B) lP'(A n B)
= lP'(B) lP'(A)lP'(B) = lP'(Ac)lP'(B).
For the final part, apply the first part to the pair B, Ac.
2. Suppose i < j and m < n. If j < m, then Aij and Amn are determined by distinct independent
rolls, and are therefore independent. For the case j = m we have that
li"(Aij n Ajn) = lP'(ith, jth, and nth rolls show Sflme number)
tlP'(jth and nth rolls both show r J ith shows r) = 3~ = li"(Aij )li"(Ajn),
r=I
as required. However, if i
i= j i= k,
3. That (a) implies (b) is trivial. Suppose then that (b) holds. Consider the outcomes numbered
i 1, i2, ... , im, and let uJ E {H, T} for 1 :::.0 j :::.0 m. Let Sj be the set of all sequences oflength M =
max{iJ : 1 :::.0 j :::.0 m} showing u1 in the i1th position. Clearly IS} I =2MI and Jn1 s1 J = 2Mm.
Therefore,
139
[1.5.4][1.7.2]
so that lP' ( nj
Solutions
Suppose IAI
a, IBI
b, lA n Bl
c, and A and B are independent. Then lP'(A n B) =
lP'(A)lP'(B), which is to say that cfp = (a/p) (b/p), and hence ab = pc. If ab =I= 0 then pI ab
(i.e., p divides ab). However, pis prime, and hence either p I a or p I b. Therefore, either A= Q or
B = Q (or both).
4.
5. (a) Flip two coins; let A be the event that the first shows H, let B be the event that the second
shows H, and let C be the event that they show the same. Then A and B are independent, but not
conditionally independent given C.
(b) Roll two dice; let A be the event that the smaller is 3, let B be the event that the larger is 6, and let
C be the event that the smaller score is no more than 3, and the larger is 4 or more. Then A and B are
conditionally independent given C, but not independent.
(c) The definitions are equivalent if 11"(C) = 1.
6.
(fo)7 < ~
k i
7. (a) lP'(A n B) = =
~ = li"(A)li"(B), and lP'(B n C) = = ~ = lP'(B)lP'(C).
(b) lP'(A n C)= 0 =/= lP'(A)lP'(C).
(c) Only in the trivial cases when children are either almost surely boys or almost surely girls.
(d) No.
8.
9.
iJ iJ =
lP'(AB I A
+ C) =
lP'(AB, A+ C)
li"(AB, B +C)
(l  p2)p2
li"(A
C)
= 1  lP'(A ~ C) = ~(=1=p=2)=2.
By a similar calculation (or otherwise) in the second case, one obtains the same answer:
2. Let A be the event of exactly one ace, and KK be the event of exactly two kings. Then lP'(A I
KK) = lP'(A n KK)/li"(KK). Now, by counting acceptable combinations,
140
.4
Solutions [1.7.3][1.8.2]
Problems
3. First method: Suppose that the coin is being tossed by a special machine which is not switched
off when the walker is absorbed. If the machine ever produces N heads in succession, then either the
game finishes at this point or it is already over. From Exercise (1.3.2), such a sequence of N heads
must (with probability one) occur sooner or later.
Alternative method: Write down the difference equations for Pk the probability the game finishes at
0 having started at k, and for /h, the corresponding probability that the game finishes at N; actually
these two difference equations are the same, but the respective boundary conditions are different.
Solve these equations and add their solutions to obtain the total 1.
4. It is a tricky question. One of the present authors is in agreement, since if li"(A
and li"(A I cc) > li"(B I cc) then
li"(A) = li"(A
> li"(B
I C)
> li"(B
I C)
= li"(B).
The other author is more suspicious of the question, and points out that there is a difficulty arising
from the use of the word 'you'. In Example (1.7.10), Simpson's paradox, whilst drug I is preferable
to drug II for both males and females, it is drug II that wins overall.
5.
< r < k)

li"(Lk=m)
li"(Lk > Lr for 1 :S r < k)
1/1 =
=
kjm.
fs.
! g= ft,.
li"(S divisible by 3)
12)
+ ll"(2, 1)}
j~
2.
[1.8.3][1.8.9]
Solutions
G)
= 1
ni A;
= 1  zn  ( ~) zn.
A7t,
={a, b, c},
n},
Then Jf = J"U g, is given by Jf = {{a}, {c}, {a, b}, {b, c}, 0, Q }. Note that {a}
but the union {a, c} is not in Jf, which is therefore not a afield.
Jf and {c}
Jf,
4. In each case J"may be taken to be the set of all subsets of Q, and the probability of any member
of J" is the sum of the probabilities of the elements therein.
(a) Q = {H, T} 3 , the set of all triples of heads (H) and tails (T). With the usual assumption of
independence, the probability of any given triple containing h heads and t = 3 h tails is ph (1 p)t,
where p is the probability of heads on each throw.
(b) In the obvious notation, Q = {U, V} 2 = {UU, VV, UV, VU}. Also JP(UU) = JP(VV) = ~ ~ and
lP(UV) = JP(VU) = ~ ~.
(c) Q is the set of finite sequences of tails followed by a head, {Tn H : n ::: 0}, together with the infinite
sequence T 00 of tails. Now, lP(TnH) = (1 p)n p, and JP(T00 ) = limn+ooO p)n = 0 if p =j:. 0.
5.
As usual, JP(A!::.. B) = 1P( (AU B)\ JP(A n B)) = JP(A U B) JP(A n B).
6.
142
Solutions
Problems
[1.8.10][1.8.13]
I C)
Q(A
Q(A n C)
lP(A n C I B)
lP(A n B n C)
Q(C) =
lP(C I B) =
lP(B n C) = JP(A
I B n C).
10. As usual,
00
00
00
= L JP(A
1
11. The first inequality is trivially true if n = 1. Let m :::: 1 and assume that the inequality holds for
n:::; m. Then
~ lP(Ai),
by the hypothesis. The result follows by induction. Secondly, by the first part,
r)
1 lP(yAf)
(n) 
= 1 n + LlP(Ai) +
LlP(Ai U Aj)0
13. Clearly,
JP(Nk)
2::
JP(
s, we write As =
n n
Ai
i ES
jrf_S
(n)
3
l<j
= (1 l)n +
by Exercise (1.3.4)
l<j
Ai)
n n
Ai
iES
+ ...
A.i).
jrf_S
ni ES Ai Then
0
143
j <k
j,kS
lP(Asu{j,kJ)
[1.8.14][1.8.16]
Solutions
where a typical summation is over all subsets S of {1, 2, ... , n} having the required cardinality.
Let Ai be the event that a copy of the i th bust is obtained. Then, by symmetry,
where aj is the probability that the j most recent ViceChancellors are obtained. Now a3 is given in
Exercise (1.3.4), and a4 and as may be calculated similarly.
I B)
lP'(A
= lP'(Aj
I Aj)lP'(Aj)
1P'(B n (U7 Ai))
JP'(B
n B)
lP'(B)
s=
4 ) = lP'({N = 2} n {S = 4}) =
JP'(S = 4)
Li JP'(S = 4 I N
1 1
12 . 4
(b) Secondly,
ft
+ 641 . I61
:! + ft + . . .
12 . 4
4 33 + 1
4433 .
lP'(N = 2 I S = 4, D = 1) =
JP'(N = 2, S = 4, D = 1)
JP'(S = 4 D = 1)
= 1
004
o. 6. 4 + o. 30. 8 + 64. I6
00
j= 1
j= 1
Finally, lP'(M
= r) = JP'(M:::: r) JP'(M:::: r
(r)j
j1 = r ( 1  r6
2
12
12
)1 =  r .
12 r
1).
16. (a) w E B if and only if, for all n, w E U~n Ai, which is to say that w belongs to infinitely many
of the An.
(b) w E
if and only if, for some n, w E n~n Ai, which is to say that w belongs to all but a finite
number of the An.
144
Solutions [1.8.17][1.8.20]
Problems
(c) It suffices to note that B is a countable intersection of countable unions of events, and is therefore
an event.
(d) We have that
n
00
Cn =
00
Ai s; An s;
U Ai =
Bn,
i=n
i=n
and therefore JP(Cn) :::: JP(An) :::: JP(Bn). By the continuity of probability measures (1.3.5), if Cn + C
then JP(Cn) + JP(C), and if Bn + B then JP(Bn) + JP(B). If B = C = A then
JP(A) = JP(C) :::: lim JP(An) :::: JP(C) = JP(A).
n>oo
17. If Bn and Cn are independent for all n then, using the fact that Cn s; Bn,
JP(Bn)lP(Cn) = JP(Bn
n Cn)
= JP(Cn) + JP(C)
as n+ oo,
and also JP(Bn)lP(Cn) + JP(B)JP(C) as n + oo, so that JP(C) = JP(B)JP(C), whence either JP(C) = 0
or JP(B) = 1 or both. In any case JP(B n C) = JP(B)JP(C).
If An + A then A = B = C so that JP(A) equals 0 or 1.
18. It is standard (Lemma (1.3.5)) that lP is continuous if it is countably additive. Suppose then that lP is
finitely additive and continuous. Let At, Az, ... be disjoint events. Then Uf Ai = limn;. 00 U1 Ai,
so that, by continuity and finiteadditivity,
19. The network of friendship is best represented as a square with diagonals, with the comers labelled
A, B, C, and D. Draw a diagram. Each link of the network is absent with probability p. We write EF
for the event that a typical link EF is present, and EfC for its complement. We write A ++ D for the
event that A is connected to D by present links.
n BCc)p + JP(A
++ D 1 ADc
n BC)(1 p)
(d)
(c)
(b)
(a)
20. We condition on the result of the first toss. If this is a head, then we require an odd number of
heads in the next n  1 tosses. Similarly, if the first toss is a tail, we require an even number of heads
in the next n  1 tosses. Hence
Pn = p(l Pnd
+ (1 P)Pn1
= (1 2p)Pn1
+P
+ (1 2p)BA.nl + p
145
+ B A. n.
[1.8.21][1.8.23]
Solutions
and A+ B = 1. Hence A=
1. B = 1. A.= 1 2p.
21. Let A = {run of r heads precedes run of s tails}, B = {first toss is a head}, and C = {first s tosses
are tails}. Then
where p = 1  q is the probability of heads on any single toss. Similarly JP(A I B)= prl + JP(A
Bc)(l_prl). WesolveforlP(A I B)andlP(A I Bc),andusethefactthatlP(A) =lP(A I B)p+JP(A
Bc)q, to obtain
I
I
22. (a) Since every cherry has the same chance to be this cherry, notwithstanding the fact that five
are now in the pig, the probability that the cherry in question contains a stone is
=
,fu ! .
(b) Think about it the other way round. First a random stone is removed, and then the pig chooses
his fruit. This does not change the relevant probabilities. Let C be the event that the removed cherry
contains a stone, and let P be the event that the pig gets at least one stone. Then JP(P I C) is the
probability that out of 19 cherries, 15 of which are stoned, the pig gets a stone. Therefore
JP(P
MH H j% .g..
23. Label the seats 1, 2, ... , 2n clockwise. For the sake of definiteness, we dictate that seat 1 be
occupied by a woman; this determines the sex of the occupant of every other seat. For 1 ::::: k ::::: 2n,
let Ak be the event that seats k, k + 1 are occupied by one of the couples (we identify seat 2n + 1 with
seat 1). The required probability is
Now, JP(Ai) = n (n  1) !2 / n !2 , since there are n couples who may occupy seats i and i + 1, (n  1)!
ways of distributing the remaining n  1 women, and (n  1)! ways of distributing the remaining n 1
men. Similarly, if 1 ::::: i < j ::::: 2n, then
(n2)t2
JP(AinAj)= { n(n1)
0
n! 2
iflijlfl
if li jl = 1,
n!
if it < i2 < < ik and ij+t  ij ::::: 2 for 1 ::::: j < k, and 2n +it  ik ::::: 2; otherwise this
probability is 0. Hence
n
(nk)l
lP nAj = ~)1)k
1 Sk,n
1
k=O
n.
(2n)
where Sk,n is the number of ways of choosing k nonoverlapping pairs of adjacent seats.
Finally, we calculate Sk,n Consider first the number Nk,m of ways of picking k nonoverlapping
pairs of adjacent seats from a line (rather than a circle) of m seats labelled 1, 2, ... , m. There is a oneone correspondence between the set of such arrangements and the set of (m  k)vectors containing
146
Solutions [1.8.24][1.8.27]
Problems
k 1'sand (m  2k) O's. To see this, take such an arrangement of seats, and count 0 for an unchosen
seat and 1 for a chosen pair of seats; the result is such a vector. Conversely take such a vector, read its
elements in order, and construct the arrangement of seats in which each 0 corresponds to an unchosen
seat and each 1 corresponds to a chosen pair. It follows that Nk,m = (mk'k).
Turning to Sk,n either the pair 2n, 1 is chosen or it is not. If it is chosen, we require another
k  1 pairs out of a line of 2n  2 seats. If it is not chosen, we require k pairs out of a line of 2n seats.
Therefore
Sk,n
= Nk1,2n2 + Nk,2n =
( 2nkI)
k_ l
+ (2nk)
k
=
(2nk) 2n
k
2n _ k
24. Think about the experiment as laying down the b + r balls from left to right in a random order.
The number of possible orderings equals the number of ways of placing the blue balls, namely (bt).
The number of ways of placing the balls so that the first k are blue, and the next red, is the number of
ways of placing the red balls so that the first is in position k + 1 and the remainder are amongst the
r + b  k  1 places to the right, namely (r+~=~ 1 ). The required result follows.
The probability that the last ball is red is r / (r + b), the same as the chance of being red for the
ball in any other given position in the ordering.
25. We argue by induction on the total number of balls in the urn. Let Pac be the probability that the
last ball is azure, and suppose that Pac = whenever a, c :::_ 1, a+ c ::; k. Let a and CJ be such that
a, a : :_ 1, a+ a = k + 1. Let A; be the event that i azure balls are drawn before the first carmine
ball, and let Cj be the event that j carmine balls are drawn before the first azure ball. We have, by
taking conditional probabilities and using the induction hypothesis, that
Paa = LPai,aiP'(A;)
+ LPa,ajiP'(Cj)
i=1
}=1
a1
= Po,aiP'(Aa) + Pa,oiP'(Ca) +
iL
a1
IP'(A;) +
1L
i=l
= 1.
IP'(Cj).
}=1
a
a+a
IP'(Aa) =  
a1
1
a+a1
a+l
ala!
= IP'(Ca).
(a+a)!
}=1
26. (a) If she says the ace of hearts is present, then this imparts no information about the other card,
which is equally likely to be any of the three other possibilities.
(b) In the given protocol, interchange hearts and diamonds.
27. Writing A if A tells the truth, and Ac otherwise, etc., the only outcomes consistent with D telling
the truth are ABCD, ABcccD, NBCcD, and NBcCD, with a total probability of
Likewise, the
only outcomes consistent with D lying are AcBcccDc, AcBcDc, ABc cDc, and ABCcDc, with a total
probability of ~. Writing S for the given statement, we have that
H.
IP'(D 1 S)
IP'(DnS)
IP'(D n S) + IP'(Dc n S)
147
13
liT
H+ ~
13
41
[1.8.28][1.8.33]
Solutions
#;
I S) =
25
8I
46
+ 8I
25
=
71
28. Let Br be the event that the rth vertex of a randomly selected cube is blue, and note that lP'(Br)
By Boole's inequality,
lo.
r=l
r=l
lP'(
(b) lP'(A
= 0.
30. The number of possible combinations of birthdays of m people is 365m; the number of combinations of different birthdays is 365!/(365 m)!. Use your calculator for the final part.
ez
32. In the obvious notation, lP'(wS, xH, yD, zC) = (!;) (~) (~) 3)
tor. Turning to the 'shape vector' (w, x, y, z) with w:::: x:::: y:::: z,
lP'(w, x, y, z) = {
I m).
if w
=f. x =
if w
= x =f. y =f. z,
148
el).
y = z,
Solutions
Problems
[1.8.34][1.8.36]
6! 5!
3! (2!) 3 ,
6!,
6!5!
4!3!2!'
6!5!
2!(3!) 2 '
6!5!
(4!)2,
6!5!
(5!)2.
35. Let Sr denote the event that you receive r similar answers, and T the event that they are correct.
Denote the event that your interlocutor is a tourist by V. Then T n vc = 0, and
lP(T
JP(T I Sr)
Hence:
= i x VI = i
I S2) = (i) 2 V[{ (i) 2 +
n V n Sr)
lP(Sr)
lP(T
n Sr I V)lP(V)
lP(Sr)
(i) 3 V[{(i) 3 +
I S4) = (i) 4 . V[{(i) 4 +
H+ ~] = i
(!) 2
(!) 3 }i +
~] =
zt
(d) JP(T
(!) 4 H + ~] = ~
(e) If the last answer differs, then the speaker is surely a tourist, so the required probability is
( 43 )3 . 41
( 1)3 X 1
4
4
+ (1)3
l
4 "4
9
= 10
36. Let E (respectively W) denote the event that the answer East (respectively West) is given.
(a) Using conditional probability,
JP(East correct I E)
JP(East correct I W)
E(6
E(~.! + ~)
= E
1
2 3
+ 3) + 34(l E)
149
E ~.
1
2 1
1
zE + (34 + 3)(1 +E)
= E,
[1.8.37][1.8.39]
Solutions
andforWWW,
E{ (~)d)3
liE
E[(~)(~)3+j]+(1E)~(i)3
Then forE =
=9+2E.
irJ, the first is !Jz; the second is i, as you would expect if you look at Problem (1.8.35).
37. Use induction. The inductive step employs Hoole's inequality and the fact that
IP'(
r=l
r=t
IP'(Ar nAt).
2~r~n
There is nothing special about the choice of At in this inequality, which will therefore hold with any
suffix k playing the role of the suffix 1. Kounias's inequality is then implied.
The above inequality holds trivially when n = 1. Assume that it holds for some value of n (2: 1).
We have that
::: :t!P'(Ar)
r=t
n+l
::: L
IP'(Ar)
r=l
Ar)
r=t
2~r~n
IP'(Ar nAt)
2~r~n+t
Therefore ak =
1
(1
nk+1
150
2::: k < n.
2
Random variables and their distributions
~.then
{w: aX(w)
{w: aX(w)
~ x} =
x}
= {w:
U1 { w: X(w) ~ ~ ~}
n2'::
{w: aX(w)
~ x} =
= 0,
if X< 0,
ifx 2: 0;
if a> 0,
if a < 0.
~
y.
3. Assume that any specified sequence of heads and tails with length n has probability zn. There
are exactly (k) such sequences with k heads.
If heads occurs with probability p then, assuming the independence of outcomes, the probability of
any givensequenceofk headsandnk tails is pk (1 p)nk. The answer is therefore (k) pk(l p)nk.
151
[2.2.1][2.4.1]
Solutions
Clearly Hn
+ Tn
lP' ( 2p  1 
as n
+
E~ ~ (Hn 
Tn)
~ 2p  1 + E) = lP' (I~ Hn 
pI ~ n
+
3. Let/n (x) be the indicator function of the event {Xn ~ x}. By the law of averages, n I I:~= 1 I r (x)
converges in the sense of (2.2.1) and (2.2.6) to lP'(Xn ~ x) = F(x).
IF(am) F(amdl
IF(x
for x E [amI, am). Hence G(x) approaches F(x) for any x at which F is continuous.
2.
3.
Certainly Y is a random variable, using the result of the previous exercise (2). Also
:F.
as required. IfF is discontinuous then p 1 (x) is not defined for all x, so that Y is not well defined.
If F is nondecreasing and continuous, but not strictly increasing, then pI (x) is not always defined
uniquely. Such difficulties may be circumvented by defining pI (x) = inf {y : F (y) ~ x}.
4. The function A.f +(1 A. )g is nonnegative and integrable over llHo 1. Finally, f g is not necessarily
a density, though it may be: e.g., iff= g = 1, 0 ~ x ~ 1 then f(x)g(x) = 1, 0 ~ x ~ 1.
5. (a) If d > 1, then J100 cxd dx = c/(d 1). Therefore f is a density function if c = d 1, and
F(x) = 1 x(dI) when this holds. If d ~ 1, then f has infinite integral and cannot therefore be a
density function.
(b) By differentiating F(x) =ex j(l +ex), we see that F is the distribution function, and c = 1.
(a) If y ~ 0,
lP'(X 2 ~ y) = lP'(X ~
X~
0. If y
,JY) lP'(X
~
<
0,
152
,JY) =
F(,J'Y) F(,J'Y).
Solutions [2.4.2][2.5.6]
Random vectors
(c)If1
1,
00
lP'(sinX ~ y) =
IP'((2n
n=oo
00
~.
a~
oo, b
oo.
r
(b) Secondly,
1 p
!x,z(x, z) =
otherwise.
3.
4.
Let A = {X
+ y 2 )}, x
::: 0, y E R.
now lP'(A n B)= lP'(A) + lP'(B) lP'(A U B), which gives the answer. Draw a map of~2 and plot the
regions of values of (X, Y) involved.
5.
Secondly, for 1
y) lP'(X = x, Y
y 1) = lP'(X = x, Y = y).
x ~ y ~ 6,
if X < y,
ifx = y.
6.
153
[2.7.1][2.7.6]
Solutions
2. 7 Solutions to problems
1.
Hence
JP>(X < x) = {
1(1p)lxJ
if X::: 0,
if X< 0.
(c) Suppose {Xm} is a sequence of random variables such that Xm(w) t X(w) for all w. Then
{X :::: X} = nm {Xm :::: X}, which is a countable intersection of events and therefore lies in :F.
3.
{X+Y::::x}=n
n=1
({X::::r}n{Y::::xr+n 1 })
rEI()>+
({X < 0}
n {Y
n {Y
< 0}).
4.
(c)
JP>(X 2 ::::
X)= JP>(X:::: 1) =
1) = l
i) = JP>(X:::: i) = !
2 ::::
5.
!P
6. There are 6 intervals of 5 minutes, preceding the arrival times of buses. Each such interval has
5 = 1 , so the answer IS
. 6 12
1 = 1.
probab1"l"1ty 60
12
2
154
Solutions [2.7.7][2.7.12]
Problems
7. Let T and B be the numbers of people on given typical flights of TWA and BA. From Exercise
(2.1.3),
~T~
(10)
k
10
)k (1 )20k
10
Now
lP'(TWA overbooked)= JP'(T = 10) = ( [0 ) 10 ,
lP'(BAoverbooked) = JP'(B:::: 19) = 20(fu) 19(fo) + (fu) 20,
of which the latter is the larger.
8.
Assuming the coins are fair, the chance of getting at least five heads is (~) 6 + 6(~) 6 =
9.
F(x)
l4.
if X< 0,
ifx :=:: 0.
(b) Secondly,
lP'(X :::0 x) = { O .
1  hmytx F(y)
if X < 0,
if X:::: 0.
lP'(IXI < x) = { 0

ifx:::: 0.
ytxo
ytxo
12. Let the dice show X andY. WriteS= X+ Y and fi = lP'(X = i), gi = JP'(Y = i). Assume that
lP'(S = 2) = lP'(S = 7) = lP'(S = 12) = 1\. Now
lP'(S = 2) = lP'(X = l)lP'(Y = 1) = flg1,
lP'(S = 12) = lP'(X = 6)lP'(Y = 6) = f6g6,
lP'(S = 7) :::: lP'(X = 1)lP'(Y = 6) + lP'(X = 6)lP'(Y = 1) = flg6 + f6g1.
It follows that
f1 g1 =
+ 1)
X
[2.7.13][2.7.14]
Solutions
where x = g6/ gJ. However x +xi > 1 for all x > 0, a contradiction.
13. (a) Clearly dL satisfies (i). As for (ii), suppose that dL(F, G)
= 0. Then
and
F(y) 0:: lim{G(y E) E}
E,l0
= G(y).
Now G(y) :::: G(x) if y > x; taking the limit as y ,j, x we obtain
F(x):::: limG(y):::: G(x),
y,lx
+ dL(G, H).
= k) lP'(Z = k)l
:S llP'(X
<0,
x,y > 0,
ifO ::0 x :S y,
ifO==oy==ox,
a2 F
dxdy = 1.
in0 oo inoo
0
axay
Hence F is a joint distribution function, and easy substitutions reveal the marginals:
Fx(x)
Fy(y)
= y+oo
lim F(x, y) =
1 ex,
x :::: 0,
= X+00
lim F(x, y) = 1 eY yeY,
156
y:::: 0.
Solutions [2.7.15][2.7.20]
Problems
15. Suppose that, for some i i= j, we have Pi < Pj and Bi is to the left of Bj. Write m for the
position of Bi and r for the position of Bj, and consider the effect of interchanging Bi and Bj. For
k :::;: m and k > r, lP'(T :::: k) is unchanged by the move. Form < k :::;: r, lP'(T :::: k) is decreased by an
amount p j  Pi, since this is the increased probability that the search is successful at the m th position.
Therefore the interchange of Bi and Bj is desirable.
It follows that the only ordering in which lP'(T :::: k) can be reduced for no k is that ordering in
which the books appear in decreasing order of probability. In the event of ties, it is of no importance
how the tied books are placed.
16. Intuitively, it may seem better to go first since the first person has greater choice. This conclusion
is in fact false. Denote the coins by C1, C2, C3 in order, and suppose you go second. If your opponent
chooses C1 then you choose C3, because 1P'(C3 beats C1) = ~ + ~ ~ = i~ >
Likewise
1P'(C1 beats C2) = 1P'(C2 beats C3) = ~ >
Whichever coin your opponent picks, you can arrange
to have a better than evens chance of winning.
17. Various difficulties arise in sequential decision theory, even in simple problems such as this one.
The following simple argument yields the optimal policy. Suppose that you have made a unsuccessful
searches "ahead" and b unsuccessful searches "behind" (if any of these searches were successful, then
there is no further problem). Let A be the event that the correct direction is ahead. Then
lP'(A
(1 p)aa
which exceeds if and only if (1  p)aa > (1  p)b(l a). The optimal policy is to compare
(1  p)aa with (1  p)b(1 a). You search ahead if the former is larger and behind otherwise; in
the event of a tie, do either.
i)
i).
18. (a) There are (6 possible layouts, of which 8+8+2 are linear. The answer is 18/(6
(b) Each row and column must contain exactly one pawn. There are 8 possible positions in the first
row. Having chosen which of these is occupied, there are 7 positions in the second row which are
admissible, 6 in the third, and so one. The answer is 8!/(<t).
19. (a) The density function is f(x)
F 1 (x)
= 2xex 2 ,
x:::: 0.
V)
= {
fu,v(u, v) du dv
= 0.
J{(u,v):u=v)
The random variables X, Y are continuous but not jointly continuous: there exists no integrable
function f : [0, 1]2 + lR such that
lP'(X :5 x, Y :5 y) =
1x 1Y
f(u, v)dudv,
u=O v=O
157
0:5 x, y :51.
3
Discrete random variables
c 1 = I::i"' 2k = 1.
(b) c = I::i"' rk fk = Iog2.
(c) c 1 = I::i"' k 2 = :rr 2 /6.
(d) c 1 = I::i"' 2k /k! = e 2  1.
1.
(a)
so the answer is
! (1 
e 2 ).
3. The number X of heads on the second round is the same as if we toss all the coins twice and count
the number which show heads on both occasions. Each coin shows heads twice with probability p 2 ,
so JP>(X = k) = (~)p2k(1 _ p2)nk.
4.
Let Dk be the number of digits (to base 10) in the integer k. Then
5. (a) The assertion follows for the binomial distribution because k(n  k) :::: (n  k
The Poisson case is trivial.
(b) This follows from the fact that k8 2: (k2  1) 4 .
(c) Thegeometricmassfunctionf(k) =qpk, k 2:0.
We have that
JP>(X
+ l)(k + 1).
Solutions
Independence
[3.2.2][3.2.3]
This, together with three similar equations, shows that X and Z are independent. Likewise, Y and Z
are independent. However
lP'(X = 1, Y = 1, Z = 1) = 0 ~ ~ = lP'(X = 1)1P'(Y = 1)1P'(Z = 1),
so that X, Y, and Z are not independent.
2.
(a) If x :::: 1,
JP>( min{X, Y} ~ x) = 1 JP'(X > x, Y > x) = 1 lP'(X > x)lP'(Y > x)
= 1 2x . 2x = 1 4x.
Since
(d)
00
00
(e)
00
00
~~zkxzx
L..., L...,
k=1x=1
L..., 2k+ 1 
k=1
3.
i<j<k
= L:o  P1)0
. 1
. 1
 pz)p~ P~ P~
i<j
1 P2P3
159
[3.2.4][3.2.5]
Solutions
(1  pl)(l  P2)
(1 P2P3)(1  P1P2P3).
4. (a) Either substitute P1 = P2 = P3 = ~ in the result of Exercise (3b), or argue as follows, with
the obvious notation. The event {A < B < C} occurs only if one of the following occurs on the first
round:
(i) A and B both rolled 6,
(ii) A rolled 6, B and C did not,
(iii) none rolled 6.
Hence, using conditional probabilities,
In calculating lP'(B < C) we may ignore A's rolls, and an argument similar to the above tells us that
1
lP'(B <C)= ( 65)2 lP'(B <C)+ 6
00
})
(~)k 1 g=
k=1,4,7, ...
Once A has thrown the first 6, the game restarts with the players rolling in order BCABCA .... Hence
the probability that B rolls the next 6 is ~~ also, and similarly for the probability that C throws the
third 6. The answer is therefore ( ~~) 3 .
5. The vector ( Xr : 1 ,::: r ,::: n) has the same joint distribution as (Xr : 1 ,::: r ,::: n), and the claim
follows.
Let X+ 2 and Y
(! Y)
6
1,::: i,j.::: 3.
1
12
Then
lP'(X = 1) = lP'(X = 1) = lP'(Y = 1) = lP'(Y = 1) = ~.
lP'(X + Y = 2) =
g # /2 = lP'(X +
160
lP'(X = 0) = lP'(Y = 0) = ~.
Y = 2).
Expectation
Solutions [3.3.1][3.3.3]
(a) No!
f(1)
lE (X)  91
= ~, f(~:) = ~,
+ 92 + 98 
f(2) =~Then
1   91
+ 98 + 92
 lE( 1/ X )
2.
(a) If you have already j distinct types of object, the probability that the next packet contains
a different type is (c j)jc, and the probability that it does not is jjc. Hence the number of days
required has the geometric distribution with parameter (c j)jc; this distribution has mean cj(c j).
(b) The time required to collect all the types is the sum of the successive times to collect each new
type. The mean is therefore
c
c1
L:c.=cL~
j=O c 1
3.
k=1
(a) Let lij be the indicator function of the event that players i and j throw the same number. Then
JE(/ij) = IJI'(lij =
o=
L: (~) 2 =
~,
:~
j.
i=1
L
lE(lij) = ~ (n).
. .
6 2
I<]
We claitn that the family {lij : i < j} is pairwise independent. The crucial calculation for this is
as follows: if i < j < k then
6
Hence
var(S) = var (
~ lij) = ~ var(lij) =
I<]
(;) var(/12)
I<]
~~=
7
12 (;).
The Xij are not pairwise independent, and you have to slog it out thus:
161
[3.3.4][3.4.1]
Solutions
4. The expected reward is I;~ 1 2k 2k = oo. If your utility function is u, then your 'fair' entrance
fee is I;~ 1 2ku(2k). For example, if u(k) = c(l  ka) fork:::: 1, where c, a > 0, then the fair
fee is
c ~ 2k(l rak) = c
L
(1 1 ).
k=1
2a+1 _ 1
This fee is certainly not 'fair' for the person offering the wager, unless possibly he is a noted philanthropist.
5.
We have that lE(Xa) = I;~ 1 xa f{x(x + 1)}, which is finite if and only if a < 1.
6.
Clearly
var(a +X)= lE( {(a+ X) JE(a +X) } 2 ) = lE({X JE(X)} 2 ) = var(X).
7. For each r, bet {1 + n(r)} 1 on horse r. If the rth horse wins, your payoff is {n(r) + 1}{1 +
n(r)} 1 = 1, which is in excess of your total stake I;k{n(k) + 1} 1.
8. We may assume that: (a) after any given roll of the die, your decision whether or not to stop
depends only on the value V of the current roll; (b) if it is optimal to stop for V = r, then it is also
optimal to stop when V > r.
Consider the strategy: stop the first time that the die shows r or greater. Let S (r) be the expected
score achieved by following this stategy. By elementary calculations,
S(6) = 6 IP'(6 appears before 1) + 1 IP'(l appears before 6) =
i,
l,
and similarly S(5) = 4, S(4) = 4, S(3) = 1 S(2) = ~ The optimal strategy is therefore to stop at
the first throw showing 4, 5, or 6. Similar arguments may be used to show that 'stop at 5 or 6' is the
rule to maximize the expected squared score.
9.
Proceeding as in Exercise (8), we find the expected returns for the same strategies to be:
S(6) = ~  3c,
S(5) = 4 2c,
S(4) = 4 ~c,
S(3) =
.!_f ~c,
S(2) = ~ c.
If c = ~,it is best to stop when the score is at least 4; if c = 1, you should stop when the score is at
least 3. The respective expected scores are ~ and
Let Ij be the indicator function ofthe event that the outcome of the (j + l)th toss is different from
the outcome of the jth toss. The number R of distinct runs is given by R = 1 +
n1
~ Ij)
2}
]=1
+ { (n 1) 2  (n 1) 2(n 2) }lE(/1) 2 .
162
Solutions
[3.4.2][3.4.5]
2. The required total is T = I:f=t X;, where X; is the number shown on the ith ball. Hence
JE(T) = klE(X t) = ~k(n + 1). Now calculate, boringly,
lE{ (
2
X;) } = klE(Xf) + k(k 1)lE(XtX2)
l=l
k ~ 2
=  L...,l
n 1
=
k(k 1)2"' ..
+ n(n
L...,l]
 1) . .
l>j
~ { 1n(n +
1)(n + 2) 
~n(n +
1)}
k(k 1) n
n(n1)~j{n(n+1)j(j+l)}
J=l
= ik(n + 1)(2n + 1) +
Hence
var(T) = k(n +
3.
(1  ~)
(1  _
m) .
2n
2n1
4. Any given red ball is in urn R after stage k if and only if it has been selected an even number of
times. The probability of this is
and the mean number of such red balls is n times this probability.
5.
Label the edges and vertices as in Figure 3.1. The structure function is
;(X)= X5 + (1 Xs){ (1 X1)X4[X3 + (1 X3)X2X6]
+Xt [x2 + (1 X2)(X3(X6 + X4(1 X6)))J}.
163
[3.4.6][3.4.9]
Solutions
6. The structure function is Irs:o:k) the indicator function of {S ::: k} where S = 2:~= 1 Xc. The
reliability is therefore 2:7=k (7) pi (1  p )ni.
7. Independently colour each vertex livid or bronze with probability ~ each, and let L be the random
set of livid vertices. Then ENL = ~ IE 1. There must exist one or more possible values of N L which
are at least as large as its mean.
8.
Let Ir be the indicator function that the rth pair have opposite polarity, so that X = 1 + 2:~~{ Ir.
t(n1).
9.
(a) Let Ai be the event that the integer i remains in the ith position. Then
= n 1 
(n)
2
1
++(1) n1 1.
n(n l)
n!
+ + (1)
n 1
.
n!
Now
lP'(M = r) =
=
(b)
n+1
dn+1 =
r=2
+ ndn,
164
Solutions
Dependence
[3.5.1][3.6.2]
where#A denotes the cardinality of the setA. By rearrangement, dn+l (n+ 1)dn = (dn ndn1).
Set Un = dn  ndn1 and note that u2 = 1, to obtain Un = (l)n, n 2: 2, and hence
n!
n!
dn =   2!
3!
+ .. + (1) nn!
.
n!
2.
oo
~ IJI'(H
= x
oo (
I N = n)IJI'(N = n) = ~ : px (1  p)nx
+
An
J,.
x!
n=x
(n  x)!
The last summation equals 1, since it is the sum of the values of the Poisson mass function with
parameter A(l  p).
3.
4. The probability of a marked animal in the nth place is ajb. Conditional on this event, the chance
of n  1 preceding places containing m  1 marked and n  m unmarked animals is
a1) (ba) /(b1),
( m1
nm
n1
as required. Now let Xj be the number of unmarked animals between the j  1th and jth marked
animals, if all were caught. By symmetry, lEXj = (b a)j(a + 1), whence lEX= m(lEXt + 1) =
m(b + 1)/(a + 1).
JR. and
= a 2 var(X)
= ( a)var(X)
+ b)var(Y) ) 2
where we have used the CauchySchwarz inequality (3.6.9) applied to X  JE(X), Y  JE(Y).
2. Let N; be the number of times the ith outcome occurs. Then N; has the binomial distribution
with parameters n and Pi.
165
[3.6.3][3.6.8]
3.
Solutions
For x = 1, 2, ... ,
00
=,?;
C{
(x + y l)(x + y)  (x + y)(x + y + 1)
C
2x(x + 1)
C(l
~
=2
1)
 x +1
'
and hence C = 2. Clearly Y has the same mass function. Finally E(X) = I;~ 1 (x + 1) 1 = oo, so
the covariance does not exist.
4.
Y)l
= 1 + h/(22p)(2+2p) = 1 +
p,
5.
(a) logy
E(1og fx(X)
fy (X)) < E[fy (X)  1] = 0
fx(X)
'
with equality if and only iffy = fx.
(b) This holds likewise, with equality if and only if f (x, y) = fx (x) fy (y) for all x, y, which is to
say that X and Y are independent.
7.
We have for 1
fx(x) =
x ~ 9 that
y=O
log (1 +
1
lOx+ y
) =log
IT (1
y=O
1
lOx+ y
8.
00
(ii) 1 =
. ak
.,a lk'

J.
kai ak}
+k' .,
. J.
ea(j + a)ai
= c,.,
.
J.
(ii) fx+r(r) =
r
crar
crar2r
L
.,
(
_
.),
=   ,, r ~ 1.
j=O J. r
J .
r.
166
) =log (1 +
_!_).
x
Solutions [3.7.1][3.7.4]
~ cr(r 1)(2a)r
1
(iii) E(X + Y 1) = L...J
= 2a. Now E(X) = E(Y), and therefore E(X) =a+ z
1
r.
r= 1
I X=
x)
y,z
=a LYIP'(Y = y, Z = z
I X=
I X=
x) + b EzJP>(Y = y, Z = z
x)
y,z
y,z
=aEyJP>(Y=y
I X=x)+bEzJP>(Z=z I X=x).
E{E(Y
IX
= LY JP>(Y = y
= X) = E(Y
IX
= X)
= m:{ E(Y I X) I X= x, Z = z }.
by part (e).
2. If and 1{1 are two such functions then E( (c/J(X) l{I(X))g(X)) = 0 for any suitable g. Setting
g(X) = l(X=x! for any x E lR such that JP>(X = x) > 0, we obtain (x) = l{l(x). Therefore
JP>(cp(X) = l{I(X)) = 1.
3. We do not seriously expect you to want to do this one. However, if you insist, the method is to
check in each case that both sides satisfy the appropriate definition, and then to appeal to uniqueness,
deducing that the sides are almost surely equal (see Williams 1991, p. 88).
4.
Now,
var(Y)
= E({Y EY}2) = E
= E(var(Y
I X))
+ var (E(Y
X)]
I X))
since the mean ofE(Y I X) is EY, and the cross product is, by Exercise (le),
=2m:[ {E(Y
167
X)
x}] = o
[3.7.5][3.7.10]
Solutions
5.
We have that
JE(T  t I T > t) =
00
Nt N t  r
(a)
JE(T t I T > t) =
N t
= ~(N t
+ 1).
r=O
oo 2(t+r)
(b)
lE(T  t I T > t) =
2=t
= 2.
r=O
6.
Clearly
lE(SIN=n)=lE(txi) =Jm,
1=1
lE(~tN).
7. A robot passed is in fact faulty with probability n = {(1  8)}/(1 8). Thus the number of
faulty passed robots, given Y, is bin(n Y, n), with mean (n Y){(1  8)}/(1  8). Hence
lE(X I Y) = Y
(n  Y).4.(1  8)
'+'
.
18
8. (a) Let m be the family size, 4>r the indicator that the rth child is female, and Jtr the indicator of
a male. The numbers G, B of girls and boys satisfy
m
B =
Jtr,
lE(G) = ~m = JE(B).
r=1
(It will be shown later that the result remains true for random m under reasonable conditions.) We
have not used the property of independence.
(b) With M the event that the selected child is male,
lE(G I M) = lE (
m1
4>r
= ~(m 1) = JE(B).
r=1
9. Conditional expectation is defined in terms of the conditional distribution, so the first step is not
justified. Even if this step were accepted, the second equality is generally false.
10. By conditioning on Xn1
lEXn = lE[lE(Xn I Xn1)] = lE[p(Xn1
168
+ lEXn1)/p.
Solutions
[3.8.1)[3.8.5)
= z) = LlP'(X = k)lP'(Y = z k)
k
k+1
if 0 ::=: k ::=: m 1\ n,
(m + 1)(n + 1)
+1
+ 1)(n + 1)
(m 1\ n)
(m
if m 1\ n < k < m v n,
m+n+1k
if m v n ::=: k ::=: m + n,
(m + 1)(n + 1)
where m 1\ n
2.
If z:::: 2,
lP'(X
L lP'(X = k, Y = z k) = z(z c+ 1) .
00
+ Y = z) =
k=1
Also, if z
:::: 0,
00
lP'(X Y
= z) = L
= k + z, Y = k)
lP'(X
k=1
00
+ z 1)(2k + z)(2k + z + 1)
00
= ic {;
,;C
(2k + z 1)(2k + z) 
oo
(2k + z)(2k + z + 1)
(l)r+1
~ (r + z)(r + z +
1)
z1
La(lal1{3(1{J)zr1=
~1
aR{(1
P
(1
R)Z1
p
lzl).
a)z1}

a{J
4. Repeatedly flip a coin that shows heads with probability p. Let Xr be the number of flips after
the r 1th head up to, and including, the rth. Then Xr is geometric with parameter p. The number
of flips Z to obtain n heads is negative binomial, and Z = I:~= 1 X r by construction.
5. Sam. Let Xn be the number of sixes shown by 6n dice, so that Xn+1
same distribution as X 1 and is independent of X n. Then,
= Xn +
lP'(Xn+1
::=:::
+ 1) = LlP'(Xn
::=:::
+ 1 r)lP'(Y = r)
r=O
6
= lP'(Xn
::=:::
n)
+ L[lP'(Xn
::=:::
+ 1 r) lP'(Xn
::=:::
n)]lP'(Y
= r).
r=O
We set g(k) = lP'(Xn = k) and use the fact, easily proved, that g(n) :::: g(n 1) :::: :::: g(n 5) to
find that the last sum is no bigger than
6
g(n) L ( r  1)1P'(Y = r) = g(n)(lE(Y) 1).
r=O
169
[3.8.6][3.9.2]
Solutions
6.
oo
( ) J..
(i)LHS='2:ng(n)eJ..A.nfn!=A.L~ne
1 A.n
1 =RHS.
n=1 n  1).
n=O
L (:J..A.n
_1)! 1xlE (g (n1
L Xr +x )) dF(x)
n
=A.
r=1
2.
= lE(T I So = k, W)
= lE(T I So= k, W, X 1 = 1)1P'(X 1 = 1 I So= k, W)
+ lE(T I So = k, W, X 1 = 1)1P'(X 1 = 1 I So =
(1 + h+ 1) Pk+1P + (1 + lkI) ( 1  Pk+1P)
Pk
= 1
+ PPk+1Jk+1 +
Pk
k, W)
Pk
as required.
Certainly Jo = 0. If p =
+ (N k + 1)h1
= 2(k N)
+ Uk1
= 2(k N),
with general solution Uk =A+ Bk 1(N k) 3 for constants A and B. Now uo =UN = 0, and
therefore A= 1N3 , B
170
Solutions [3.9.3][3.10.3]
3. The recurrence relation may be established as in Exercise (2). Set uk = (pk pN)h and use
the fact that Pk = (pk pN)/(1 pN) where p = q jp, to obtain
PUk+1  (1 r)uk
The solution is
Uk
4.
Conditioning in the obvious way on the result of the first toss, we obtain
Pmn
ifm,n:::: 1.
Let Y be the number of negative steps of the walk up to absorption. Then lE(X + Y) = Dk and
N  k if the walk is absorbed at N,
XY= {
k
Hence lE(X  Y) = (N  k)(1  Pk)  kpk. and solving for lEX gives the result.
I x1
= !f1(2n 1)
= 1)
+ !lP'(T =
2n
I x1
= 1)
+ !f1(2n 1)
where fb(m) is the probability that the first passage to b of a symmetric walk, starting from 0, takes
place at time m. From the hitting time theorem (3.10.14),
2n1
I:f
2.
r:::: 0,
+ 1)
+ 1) + lP'(Sn = r) 2lP'(Sn :::: r + 2) lP'(Sn = r + 1)
= lP'(Sn = r) + lP'(Sn = r + 1)
= max{lP'(Sn = r), lP'(Sn = r + 1)}
= 2lP'(Sn :::: r
3.
from the arc sine law (3.10.19) for the last visit to the origin.
171
[3.11.1][3.11.4]
Solutions
1.
lP'(X=x,Y=y)
x,y:
g(x )=a,h(y )=b
lP'(X
= x)lP'(Y = y)
x,y:
g(x )=a,h(y )=b
lP'(X = x)
L
x:g(x)=a
lP'(Y = y)
L
y:h(y)=b
fy(y)
=L
fx.r(x, y)
= h(y) Lg(x).
Now
so that
fx(x)fy(y)
= g(x)h(y) L
g(x) L
h(y)
3.
= x) = 0 for x
= 0) = 1.
= 0) =
1.
(a)
E(g(X))
= LYlP'(g(X) = y) = L
Y
ylP'(X
Y x:g(x)=y
= x) = Lg(x)lP'(X = x)
x
as required.
(b)
E(g(X)h(Y))
=L
g(x)h(y)fx.r(x, y)
by Lemrna(3.6.6)
x,y
= Lg(x)h(y)fx(x)fy(y)
by independence
x,y
= fy(i) = ~fori = 1, 2, 3.
(b) (X+ Y)(w,) = 3, (X+ Y)(wz) = 5, (X+ Y)(w3) = 4, and therefore fx+Y(i) = ~fori = 3, 4, 5.
(c) (XY)(w!) = 2, (XY)(wz) = 6, (XY)(w3) = 3, and therefore fxy(i) = ~fori = 2, 3, 6.
4.
172
Problems
Solutions
(a)
lP'(Y = 2, Z = 2)
lP'(wJ)
lP'(Z = 2)
= lP'(wJ u W2) =
I 3) = fz1rO I 1) = 1.
1} = k, and therefore k =
L k = k n=l
L {1n  n=i n(n + 1)
n+1
00
2'
5.
i. ~. 3.
frlz(2 I 2) =
(e)
[3.11.5][3.11.8]
00
1.
(b) 2::~ 1 kna = kl;(a) where 1; is the Riemann zeta function, and we require a< 1 for the sum
to converge. In this case k = 1; (a) I .
6.
lP'(X + Y = n) =
L lP'(X = n 
k)lP'(Y = k) =
k=O
A.;..nk
eJLJtk
lP'(X = k, X+ Y = n)
lP'(X = k I X+ Y = n) = lP'(X_+_Y_=_n_)_
(b)
lP'(X=n+k,X >n)
lP'(X=n+k I X >n)=       lP'(X > n)
)n+k1
(1
P  P
.
= p(1  p)ki = lP'(X = k).
L:~n+i p(1 p)J1
(ii) Many random variables of interest are 'waiting times', i.e., the time one must wait before the
occurrence of some event A of interest. If such a time is geometric, the lackofmemory property
states that, given that A has not occurred by time n, the time to wait for A starting from n has the same
distribution as it did to start with. With sufficient imagination this can be interpreted as a failure of
memory by the process giving rise to A.
(iii) No. This is because, by the above, any such process satisfies G(k + n) = G(k)G(n) where
G(n) = lP'(X > n). Hence G(k + 1) = G(li+ 1 and X is geometric.
8.
Clearly,
k
lP'(X + y = k) = LlP'(X = k j, y = j)
j=O
=
t (:
j=O
.)Pkjqmk+j
J
m)
(~)pjqnj
J
k m+nk LJ ( k
pq
. (n)
.
j=O
 J
J
173
k m+nk
pq
(m k+ n)
[3.11.9][3.11.13]
+ n, p).
which is bin(m
9.
Solutions
as required. Now,
IP(N even) =
L (~) pk (1 _ p )nk
k even
= i{ (p + 1 
P)n
+ (1
 P  P )n}
= ~ {1 + (1 
2p )n}
(%)
ways of choosing k blue balls, and (~~f) ways of choosing n k red balls. The
total number of ways of choosing n balls is (~),and the claim follows. Finally,
IP(B
= k) =
n)
b!
(N b)!
(N n)!
k (b k)! . (N b n + k)! .
N!
=(~){~b~1b~+1}
x { N; b
~ ( ~) pk(1 
... N b;
as N
p)nk
+ k + 1} { ~ ...
N;
+ 1} I
~ oo.
12. (a) E(X) = c + d, E(Y) = b + d, and E(XY) = d, so cov(X, Y) = d (c + d)(b +d), and X
and Y are uncorrelated if and only if this equals 0.
(b) For independence, we require f (i, j) = JP( X = i )JP( Y = j) for all i, j, which is to say that
a=(a+b)(a+c),
b=(a+b)(b+d),
c=(c+d)(a+c),
d=(b+d)(c+d).
Now a+ b + c + d = 1, and with a little work one sees that any one of these relations implies the
other three. Therefore X and Yare independent if and only if d = (b + d)(c +d), the same condition
as for uncorrelatedness.
oo
E(X)
m=O
m!P(X
= m) =
L L
oo
JP(X
= m) =
m=O n=O
oo
L L
n=Om=n+!
174
oo
JP(X
= m) = LIP(X
n=O
> n).
Solutions [3.11.14][3.11.14]
Problems
(b) First method. Let N be the number of balls drawn. Then, by (a),
r
E(N) =
n=O
n=O
r
r
r 1
r n + 1
r
r!
(b + r n)!
=.L:b+rb+rlb+rn+l =.L:(b+r)! (rn)!
n=O
n=O
=
t (n
(b+r)! n=O
+b) = b + r + 1
b
b+ 1 '
+ G)
f>rt
r=O
n=O
(nbb)
rt!ti).
(n+b)
b
_ 1 f=xn(n+b)
l  X n=O
b
= (1 x)(b+2) =
f=xr
r=O
(b + ++ 1)
r
by the (negative) binomial theorem. Equating coefficients of xr, we obtain the required identity.
Second method. Writing m(b, r) for the mean in question, and conditioning on the colour of the first
ball, we find that
b
r
m(b,r) =   + {1 +m(b,r 1)}.
b+r
b+r
With appropriate boundary conditions and a little effort, one may obtain the result.
Third method. Withdraw all the balls, and let Ni be the number of red balls drawn between the i th and
(i + l)th blue ball (No = N, and Nb is defined analogously). Think of a possible 'colour sequence'
as comprising r reds, split by b blues into b + 1 red sequences. There is a oneone correspondence
between the set of such sequences with No = i, Nm = j (for given i, j, m) and the set of such
sequences with No = j, Nm = i; just interchange the 'Oth' red run with the mth red run. In particular
E(No) = E(Nm) for all m. Now No+ NI + + Nb = r, so that E(Nm) = r /(b + 1), whence the
claim is immediate.
(c) We use the notation just introduced. In addition, let Br be the number of blue balls remaining after
the removal of the last red ball. The length of the last 'colour run' is Nb + Br, only one of which is
nonzero. The answer is therefore rf(b + 1) + bf(r + 1), by the argument of the third solution to part
(b).
14. (a) We have that E(Xk) = Pk and var(Xk) = PkO Pk), and the claims follow in the usual way,
the first by the linearity of][ and the second by the independence of the Xi; see Theorems (3.3.8) and
(3.3.11).
(b) Lets = L:k Pb and let Z be a random variable taking each of the values PI, P2 ... , Pn with
equal probability nI. Now E(Z 2 )  E(Z) 2 = var(Z) 2:: 0, so that
Lk
;;Pt 2::
Lk
1
;;Pk
)2 =
s2
n2
with equality if and only if Z is (almost surely) constant, which is to say that PI = P2 = = Pn
Hence
175
[3.11.15][3.11.18]
Solutions
Hence, by the result of Problem (3.11.2), L:k Xk(Xk EXk) is constant with probability one, and the
result follows.
16. The random variables X+ Y and IX Yl are uncorrelated since
cov(X + Y, IX Yl) = E{ (X+ Y)IX Yl} E(X + Y)E(IX Yl)
=
! + !  1. ~ =
0.
However,
m:(t
h)
 1
= tE(h) 2 +
1
LE(Ijh) 1
j#
=1
+2(n)
1
1
2 n(n 1)
= 1.
1 nr (1i
lP'(X = r) =  ' """"'
L.J  z.., '
r. i=O
0 _::: r _::: n 2,
176
Solutions [3.11.19][3.11.19]
Problems
to see this consider the three cases X = X', X < X', X > X' separately, using the fact that
are increasing. Taking expectations, we obtain
and g
2{ E(f(X)g(X)) 
+ E(f(X )g(X
1
1 ))
E(f(X))E(g(X))}
Ix 1, Xz, .. ., xk1)} ;
here, the conditional expectation given X 1, Xz, ... , Xk1 is defined in very much the same way as in
Definition (3.7.3), with broadly similar properties, in particular Theorem (3.7.4); see also Exercises
(3.7.1, 3). If X 1, X 2, ... , Xk1 are given, then f(X) and g(X) may be thought of as increasing
functions of the single remaining variable Xko and therefore
E(f(X)g(X) X1, Xz, ... , Xk1) ::: m:(f(X) X1, Xz, ... , Xk1)m:(g(X) X1, Xz, ... , Xk1)
are increasing functions of the k 1 variables X 1 , X 2, ... , X k1 , implying by the induction hypothesis
that E(f' (X)g' (X)) ::: E(f' (X) )E(g' (X)). We substitute this into (*) to obtain
E(f(X)g(X)) 2:: E(f' (X))E(g' (X))= E(f(X))E(g(X))
= pN(UJ)qmN(UJ) where p + q = 1.
Applying the CauchySchwarz inequality (3.6.9) to the latter covariance, we find that R' (p) :::;
(pq) 1Jvar(IA) var(N). However lA is Bernoulli with parameter R(p), so that var(IA) = R(p)(lR(p)), and finally N is bin(m, p) so that var(N) = mp(l p), whence the upper bound for R'(p)
follows.
As for the lower bound, use the general fact that cov( X+ Y, Z) = cov( X, Z) +cov( Y, Z) to deduce
that cov(/A, N) = cov(/A, lA) + cov(IA, N lA) Now lA and N lA are increasing functions of
w, in the sense of Problem (3.11.18); you should check this. Hence cov(/A, N) 2:: var(/A) + 0 by the
result of that problem. The lower bound for R' (p) follows.
177
[3.11.20][3.11.21]
Solutions
20. (a) Let each edge be blue with probability P1 and yellow with probability P2; assume these two
events are independent of each other and of the colourings of all other edges. Call an edge green if it
is both blue and yellow, so that each edge is green with probability P1 P2 If there is a working green
connection from source to sink, then there is also a blue connection and a yellow connection. Thus
IP'(green connection) :::0 IP'(blue connection, and yellow connection)
= IP'(blue connection)IP'(yellow connection)
so that R(P1 P2) :::0 R(pt)R(p2).
(b) This is somewhat harder, and may be proved by induction on the number n of edges of G. If n = 1
then a consideration of the two possible cases yields that either R(p) = 1 for all p, or R(p) = p for
all p. In either case the required inequality holds.
Suppose then that the inequality is valid whenever n < k where k :::: 2, and consider the case
when G has kedges. Let e be an edge of G and write w(e) for the state of e; w(e) = 1 if e is working,
and w(e) = 0 otherwise. Writing A for the event that there is a working connection from source to
sink, we have that
R(pY) = IP'pr (A
+ IP'p(A
1 w(e)
= O)Y (1  pY)
where IP'a is the appropriate probability measure when each edge is working with probability a. The
inequality here is valid since, if w(e) is given, then the network G is effectively reduced in size by one
edge; the induction hypothesis is then utilized for the case n = k  1. It is a minor chore to check that
xYpY
+ yY(l p)Y
:::0 {xp
+ y(l p)JY
if X 0::: y 0::: 0;
to see this, check that equality holds when x = y :::: 0 and that the derivative of the lefthand side with
respect to x is at most the corresponding derivative of the righthand side when x, y :::: 0. Apply the
latter inequality with x = IP'p(A I w(e) = 1) andy= IP'p(A I w(e) = 0) to obtain
R(pY) :::0 {IP'p(A
R(p)Y.
21. (a) The number X of such extraordinary individuals has the bin(107 , 10 7) distribution. Hence
lEX= 1 and
1 2e 1
1 e1
:::::::
0.4.
1 2e 1
(c) Provided m
= 107,
IP'(X = m) =
N!
( 1) m (
1 ) N m
e 1
1 :::::::  ,
m!(N m)! N
N
m!
178
Solutions [3.11.22][3.11.23]
Problems
the Poisson distribution. Assume that "reasonably confident that n is all" means that lP'(X > n I X ::::
n) :::0 r for some suitable small number r. Assuming the Poisson approximation, lP'(X > n) :::0 r lP'(X ::::
n) if and only if
00
00
e 1 ~  < re 1 ~  .
~ k!~k!
k=n+1
k=n
For any given r, the smallest acceptable value of n may be determined numerically. If r is small, then
very roughly n::::::: 1/r will do (e.g., if r = 0.05 then n::::::: 20).
(d) No level p of improbability is sufficiently small for one to be sure that the person is specified
uniquely. If p = w 7 a, then X is bin(107 , w7 a), which is approximately Poisson with parameter
a. Therefore, in this case,
lP'(X > 1 I X:::: 1):::::::
1  e 01  ae 01
= p,
1 e 01
say.
An acceptable value of p for a very petty offence might be p ::::::: 0.05, in which case a : : : : 0.1 and so
p = w 8 might be an acceptable level of improbability. For a capital offence, one would normally
require a much smaller value of p. We note that the rules of evidence do not allow an overt discussion
along these lines in a court of law in the United Kingdom.
22. The number G of girls has the binomial distribution bin(2n, p). Hence
:::0 ( 2n)
n
= 0.485 and n =
(2n) p n q n  q ,
n
qp
~
~p k q 2nk =
k=n
( 2n) Pn qn _ q  :::0
q p
~ { (1 
v (nrr)
0.03)(1
+ 0.03)} n 00.50135
.
104
0.515
=  ( 1 9 )
< 1.23 x
3,Jir
lo4
w 5 .
It follows that the probability that boys outnumber girls for 82 successive years is at least (1  1.23 x
w 5 ) 82 :::: o.99899.
23. Let M be the number of such visits. If k =f. 0, then M :::: 1 if and only if the particle hits 0 before
it hits N, an event with probability 1 kN 1 by equation (1.7.7). Having hit 0, the chance of another
visit to 0 before hitting N is 1  N 1, since the particle at 0 moves immediately to 1 whence there is
probability 1  N 1 of another visit to 0 before visiting N. Hence
lP'(M :::: r I So
= k) = (1 
k) (1  N1 )r1 ,
r :::: 1,
so that
lP'(M = j
I So =
k) = lP'(M :::: j
I So =
k)  lP'(M :::: j
=(1:)(1~y1~,
179
+ 1 I So =
j::::l.
0)
[3.11.24][3.11.27]
Solutions
24. Either read the solution to Exercise (3.9 .4), or the following two related solutions neither of which
uses difference equations.
First method. Let Tk be the event that A wins and exactly k tails appear. Then k < n so that
lP'(A wins) = L:/::JlP'(Tk). However lP'(Tk) is the probability that m + k tosses yield m heads, k tails,
and the last toss is heads. Hence
m+n1 (
+;
l)
pkqm+n1k.
k=m
1 (qjp)k
1 (qjp)N
1 + (q/p)zk.
1 (q/p)zk
1 (qjp)2N
1 + (qjp)'ZN
'
see Example (3.9.6). If k and N are even, doubling the stake is equivalent to playing the original game
with initial fortune k and the price of the Jaguar set at
The probability of winning is now
iN.
lk
1 (q/p)'Z
I
1 (qjp)'ZN
which is larger than before, since the final term in the above display is greater than 1 (when p <
i,
If p =
doubling the stake makes no difference to the chance of winning. If p >
to decrease the stake.
i ).
i, it is better
26. This is equivalent to taking the limit as N ~ oo in the previous Problem (3.11.25). In the limit
when p =f.
the probability of ultimate bankruptcy is
i,
+q =
l. If p =
l.
Rn1
180
1, whence
Solutions [3.11.28][3.11.29]
Problems
since IP'(S1S2 Sm =f. 0) ,[_ IP'(Sk =f. 0 for all k 2:: 1) as m ~ oo.
There are various ways of showing that the last probability equals Ip  q I, and here is one.
Suppose p > q. If x 1 = 1, the probability of never subsequently hitting the origin equals 1 (q/p),
by the calculation in the solution to Problem (3 .11.26) above. If X 1 = 1, the probability of staying
away from the origin subsequently is 0. Hence the answer is p(1 (qfp)) + q 0 = p q.
If q > p, the same argument yields q p, and if p
=q=
~the answer is 0.
28. Consider first the event that M2n is first attained at time 2k. This event occurs if and only if: (i)
the walk makes a first passage to S2k (> 0) at time 2k, and (ii) the walk thereafter does not exceed
S2k. These two events are independent. The chance of (i) is, by reversal and symmetry,
IP'(S2k1 < s2k s2k2 < S2b ... , So < S2k)
= IP'(X2k > o, x 2k_ 1 + x 2k > o, ... , x 1 + ...
= IP'(X 1 >
= IP'(Si >
+ x 2k > O)
o, x 1 + x 2 > o, 00., x 1 + 00 . + x 2k > O)
0 for 1 :::0 i :::0 2k) = ~IP'(Si =f. 0 for 1 :::0 i :::0 2k)
= ~IP'(S2k = 0)
by equation (3.10.23).
As for the second event, we may translate S2k to the origin to obtain the probability of (ii):
IP'(S2k+1 :::0 S2b ... , S2n :::0 S2k) = IP'(M2n2k = 0) = IP'(S2n2k = 0),
where we have used the result of Exercise (3.10.2). The answer is therefore as given.
The probabilities of (i) and (ii) are unchanged in the case i = 2k + 1; the basic reason for this is
that S2r is even, and S2r+ 1 odd, for all r.
29. Let Uk = IP'(Sk = 0), fk = IP'(Sk = 0, Si
recall from equation (3.10.25)) to obtain
U2n =
L U2n2khk
k=1
Now N1 = 2, and therefore it suffices to prove that E(Nn) = E(Nn1) for n:::: 2. Let N~_ 1 be
the number of points visited by the walk S1, S2, ... , Sn exactly once (we have removed So). Then
N~_ 1 + 1 if Sk =f. So for 1 :::0 k :::0 n,
{
Nn =
N~_ 1  1 if Sk =So for exactly one kin {1, 2, 00., n},
N~_ 1
otherwise.
+ f4an4 + + hLn/2J}
where LxJ is the integer part of x. Now a2m = a2m+1 = u2m by equation (3.10.23). If n = 2k is
even, then
E(N2k) E(N2k1) = u2k {f2u2k2
If n
+ + hkl =
+ + hk} =
= 2k + 1 is odd, then
E(N2k+d E(N2k) = u2k {hu2k2
181
[3.11.30][3.11.34]
Solutions
lP'(M=m)=
( . . )= IT (1p1 )
IT lP'N(z)=m(z)
i
where
c = IJ;(l i;tJ).
Pi
1
{Jm(i)=C
Pi
c 1 =
II; pr(i)
for nonnegative
L:m mfJ.
32. Number the plates 0, 1, 2, ... , N where 0 is the starting plate, fix k satisfying 0 < k ::; N, and
let Ak be the event that plate number k is the last to be visited. In order to calculate lP'(Ak), we cut
the table open at k, and bend its outside edge into a line segment, along which the plate numbers read
k, k + 1, ... , N, 0, 1, ... , kin order. It is convenient to relabel the plates as (N + 1 k), (Nk), ... , 1, 0, 1, ... , k. Now Ak occurs if and only if a symmetric random walk, starting from 0,
visits both (N k) and k 1 before it visits either (N + 1  k) or k. Suppose it visits (N k)
before it visits k  1. The (conditional) probability that it subsequently visits k  1 before visiting
(N + 1  k) is the same as the probability that a symmetric random walk, starting from 1, hits N
before it hits 0, a probability of N 1 by (1.7.7). The same argument applies if the cake visits k 1
before it visits (N k). Therefore lP'(Ak) = N 1.
33. With j denoting the jth best vertex, the walk has transition probabilities Pjk = ( j  1) 1 for
1 ::; k < j. By conditional expectation,
rj
j1
= 1 + . .l.>b
TI = 0.
1 1 k=1
34. Let Pn denote the required probability. If (mr, mr+1) is first pair to make a dimer, then m1
is ultimately uncombined with probability Pr1 By conditioning on the first pair, we find that
Pn = (P1 + P2 + + Pnz)/(n  1), giving n(Pn+1  Pn) = (Pn  Pn1). Therefore,
n! (Pn+ 1  Pn) = (l)n 1 (pz  P1) = (l)n, and the claim follows by summing.
Finally,
n
r=1
since the rth molecule may be thought of as an end molecule of two sequences oflength r and n r + 1.
Now Pn * e 1 as n* oo, and it is an easy exercise of analysis to obtain that n 1EUn * e 2.
182
Solutions [3.11.35][3.11.36]
Problems
35. First,
where the last summation is over all subsets {r1, ... , rk} of k distinct elements of {1, 2, ... , n}.
Secondly,
Ak :S k!
:S k!
L
PqPr2 Prk
{q, ... ,rkl
L
PqPr2 Prk
r,, ... ,rk
+ (~)
+ (~)
LPf
i
mfXPi
L
Pr1Pr2 Prk2
r,, ... ,rk2
(~ Pj) k 1
J
Hence
By Taylor's theorem applied to the function log ( 1x), there exist t:lr satisfying 0 < t:lr < {2( 1c)2 )}  1
such that
n
Finally,
lP'(X = k) =
(IIr (1 
Pr ))
Pq ... Prk
1 N
1 N
n r=1
We write Y
=f.(.
n r=1
Y  E(Y)

= ~Xr
~ r=1 n
n) .
( Ir  
183
[3.11.37][3.11.38]
Solutions
HnC
c
HnC
nC
(1 y)a
1y
HnC
Figure 3.2. The tree of possibility and probability in Problem (3.11.37). The presence of the
disease is denoted by C, and hospitalization by H; their negations are denoted by C and H.
that
var(Y)
N
=L
x?:
n
n 2 E { ( Ir N
i#j
r=l
~ x?: !!_ ( 1 L
r=l
n2 N
~ 2 Nn
= Lxr N2n
r=l
)2} + L ;:;rE
XiXj { ( /iN
n) ( Ij N
n)}
!!_)
N
+""' XiXj
L n2
i#j
{!!_N Nn 1
_ !!!:.__}
 1 N2
""'
Nn
 ~XiXj n(N 1)N2
z#J
r=l
z#J
N  n 1 {
2
2}
N  n 1
_2
= n(N1)N :=txr Nx
= n(Nl)N :=t(xrx).
37. The tree in Figure 3.2 illustrates the possibilities and probabilities. If G contains n individuals,
X isbin(n, yp+ (1 y)a) and Yis bin(n, yp). Itisnotdifficultto seethatcov(X, Y) = nyp(lv)
where v = yp + (1 y)a. Also, var(Y) = nyp(1 yp) and var(X) = nk(1 v). The result follows
from the definition of correlation.
38. (a) This is an extension of Exercise (3.5.2). With lP'n denoting the probability measure conditional
on N = n, we have that
184
Solutions [3.11.39][3.11.39]
Problems
where s = n  I:~=l r;. Therefore,
00
n=O
=IT {
vri J(iY~,evf(i)}
i=l
r1
vs(l
~(kW ev(IF(k)).
s=O
00
00
00
00
39. (a) Place an absorbing barrier at a+ 1, and let Pa be the probability that the particle is absorbed
at 0. By conditioning on the first step, we obtain that
1 :::0 n :::0 a.
The boundary conditions are Po = 1, Pa+l = 0. It follows that Pn+l  Pn = (n + 1)(pn  Pnd
for 2 :::0 n :::0 a. We have also that P2  PI = PI  1, and
Pn+l Pn =
Setting n =a we obtain that Pa = iCa + 1)! (PI 1). By summing over 2 :::0 n <a,
a
(a+ 1)!
4 + 3! + 4! ++(a+ 1)!
It is now easy to see that, for given r, Pr = Pr (a) + 1 as a + oo, so that ultimate absorption
at 0 is (almost) certain, irrespective of the starting point.
(b) Let Ar be the probability that the last step is from 1 to 0, having started at r. Then
185
r::: 2.
[3.11.40][3.11.40]
Solutions
It follows that
Ar Ar1
whence
1
(Ar+1  Ar),
r+l
1
A3  A2 = 4 . 5 ... (r
+ l) ('Ar+1 
Ar ),
~ 2. From(**) with r
3.
= 2, 'A2 = 1Al>
= 0, and
1,
whence Jtr+1 Jtr = (r + l)(Jtr Jtr1) 1 for r ~ 2. Therefore, Vr+1 = (Jtr+1 Jtr)/(r
satisfies Vr+ 1  Vr = 1 / (r + 1)! for r ~ 2, and some further algebra yields the value of Jt 1
+ 1)!
40. We label the vertices 1, 2, ... , n, and we let rr be a random permutation of this set. Let K be the
set of vertices v with the property that rr(w) > rr(v) for all neighbours w of v. It is not difficult to
see that K is an independent set, whence a( G) ~ IKI. Therefore, a( G) ~ ElK I = 'L:v lP'(v E K).
For any vertex v, a random permutation rr is equally likely to assign any given ordering to the set
comprising v and its neighbours. Also, v E K if and only if v is the earliest element in this ordering,
whence lP'(v E K) = lj(dv + 1). The result follows.
186
4
Continuous random variables
(a) {x(l  x)} ~ is the derivative of sin 1(2x  1), and therefore C = n 1.
(b) C = 1, since
(c) Substitute v = (1
+ x 2 ) 1 to obtain
dx
00
oo (1+,x2=)m =
{1 m 3
I
1
1
v ~ (1 v)~ dv = B(2, m 2)
lo
where B(, )is a beta function; see paragraph (4.4.8) and Exercise (4.4.2). Hence, if m >
i,
r( 21 )r(m  21 )
C 1 _ B(l2'm _ 21) _
r(m)
2.
yfa)
Fx(yfa).
= a 1 fx(y/a).
(ii) Certainly
lP'(X :S y)
= lP'(X::: y) =
00
y
fx(x)dx
1Y
fx(u)du
00
1Y
fx(u)du
= lP'(X :S y),
00
+ (1 
187
[4.1.4][4.2.4]
Solutions
If X is a random variable with density f, and Y a random variable with density g, then a f +(1a) g
is the density of a random variable Z which takes the value X with probability a and Y otherwise.
Some minor technicalities are necessary in order to find an appropriate probability space for such
a Z. If X andY are defined on the probability space (Q, :F, lP'), it is necessary to define the product
space (Q, :F, lP') x (:E, g., Q) where :E = {0, 1}, g. is the set of all subsets of :E, and Q(O) = a,
Q(l) = 1 a. For w x a E Q x :E, we define
Z(w x a)= {
X(w)
if a= 0,
Y(w)
if a = 1.
..
. 1 F(x +h) F(x)
f(x)
(a) By defimtion, r(x) = hm= h10 h
1 F(x)
1  F(x)
(b) We have that
4.
H(x)
x
Jo
r(x)  2_
x
x2
Jo
r(y) dy
= 2_
x2
Jo
1
1
H(ax) _:::: H(x)
ax
x
for all x
0,
which is to say that a 1 log[l  F(ax)] _:::: log[l F(x)]. We exponentiate to obtain the claim.
(d) Likewise, if H(x)/x is nondecreasing then H(at) _:::: aH(t) for 0 _::::a _:::: 1 and t ~ 0, whence
H(at) + H(t at) _:::: H(t) as required.
2.
~ 1, the geometric
v) = lP'(X :S v, Y :S
v)
= lP'(X
:S v)lP'(Y :S v)
= F(v) 2 .
= 1 lP'(U
= 2f(u)[l
> u)
= 1 [1 F(u)] 2 ,
 F(u)].
3. The 24 permutations of the order statistics are equally likely by symmetry, and thus have equal
probability. Hence lP'(Xt < Xz < X3 < X4) =
and lP'(Xt > Xz < X3 < X4) =
by
enumerating the possibilities.
f4,
4. lP'(Y(y) > k)
Therefore,
:A,
F(y)/[1  F(y)]
+ oo as + oo.
188
F(y)
1  F(y)
J}+ e
as y+ oo.
Solutions [4.3.1][4.3.5]
Expectation
1.
We have that
X;)
( L.,n1 
1 = JE:
= LE(X;/Sn).
Sn
i=1
By symmetry, JE:(X;/Sn) = JE:(XlfSn) for all i, and hence 1 = nJE:(X1/Sn). Therefore
m
3.
An alternative proof is as follows. Let lx be the indicator of the event that X > x, so that
J~ lx dx =X. Taking expectations, and taking a minor liberty with the integral which may be made
rigorous, we obtain EX = J000 E(Ix) dx. A similar argument may be used for the more general case.
4. We may suppose without loss of generality that tt = 0 and a = 1. Assume further that m > 1.
In this case, at least half the probability mass lies to the right of 1, whence JE:(X/{x:::mJ) ::: ~ Now
O=E(X) =E{X[/{x:::m} +l{X<mJD,implyingthatJE:(Xl{X<mJ):::: ~.Likewise,
E(X 2 /{x:::mJ) ::: ~,
It follows that var(X I X < m):::: 0, which implies in tum that, conditional on {X < m}, X is almost
surely concentrated at a single value. This contradicts the continuity of X, and we deduce that m :::: 1.
The possibility m < 1 may be ruled out similarly, or by considering the random variable X.
5.
x+ and x
r [1F(x)]dx lor
00
lo
00
x
F(x)dx=
r [1F(x)]dxjoo F(x)dx.
00
lo
It is a triviality that
189
[4.4.1)[4.4.6)
Solutions
fooo x
r(t) =
1ex dx = (t
1
1) fooo x
1 2 ex
dx = (t l)r(t
1).
If n is an integer, then it follows that r(n) = (n 1)r(n 1) = = (n 1)! r(l) where r(l) = 1.
(ii) We have, using the substitution u2 = x, that
4Joroo
e" 2 du
roo ev
{fooo 2e" 2 du
=
2
Jo
41r=O Je=O
r1(
12 er2 r dr d() = 7r
00
dv =
2.
1
+ z)
=
1
1
1
3
1
1
(2n)! '(n z)r(n :z) = ... = (n z)(n z) ... :zr{z) = 4nn! y7r.
Now set u
oo r1
1u=Olv=O
3. If g is strictly decreasing then lP'(g(X) ::S y) = lP'(X :::: g 1 (y)) = 1 g 1 (y) so long as
0 ::s g 1(y) ::s 1. Therefore lP'(g(X) ::s y) = 1  eY, y:::: 0, if and only if g 1 (y) = eY, which is
to say that g(x) = logx for 0 < x < 1.
4.
We have that
lP'(X < x)
joo rr(1 +
x
u2 )
du
7r
=  + tan 1 x.
Also,
lE(IXIa)
= ~oo
lxla 2 dx
oorrO+x)
Writing <I> for the N (0, 1) distribution function, lP'(Y ::S y) = lP'(X ::S logy) = <I> (logy). Hence
1
1
_I (1
)2
fy(y) = fx(logy) =   e 2 ogy ,
6.
y$
0< y<
00.
Integrating by parts,
LHS =
L:
g(x) { (x
= [g(x)a
C: JL)] ~00 + L:
190
g'(x)a
C: JL)
dx = RHS.
Dependence
7.
Solutions
[4.4.7][4.5.3]
8.
1 <l>(x) =
1oo
tj>(u)du = 
= tj>(x) _ tj>(x)
x
1oo
1
x3
4>'(u)
tj>(x)
du = U
00 3'(u) du =
u5
1oo 
tj>(x) _ tj>(x)
x
x3
q/(u)
3du
U
+ 3(x)
x5
100
x
15(u) du.
u6
1oo00
I 2 2
ezx Y dy = ~elxl
J2nx2
if x =f. 0, since the integrand is the N(O, x 2 ) density function. It is easily seen that g(O) = 0, so that
g is discontinuous, while
100oo
g(x)dx =
1oo ~elxl
00
dx = 1.
00 100
1oo
oo !Q(x,y)dxdy = 2::00
n=l
Ur
.1
1.
Also f Q is the uniform limit of continuous functions on any subset of ~2 of the form [ M, M] x ~;
hence f Q is continuous. Hence f Q is a continuous density function. On the other hand
00
oo
!Q(x, y)dy =
00
2:: Or g(x qn).
n=l
where g is discontinuous at 0.
(iii) Take Q to be the set of the rationals, in some order.
2. We may assume that the centre of the rod is uniformly positioned in a square of size a x b, while
the acute angle between the rod and a line of the first grid is uniform on [0, ~ n]. If the latter angle is
e then, with the aid of a diagram, one finds that there is no intersection if and only if the centre of the
rod lies within a certain inner rectangle of size (a r cos&) x (b r sin&). Hence the probability of
an intersection is
7r~
3. (i) Let I be the indicator of the event that the first needle intersects a line, and let J be the indicator
that the second needle intersects a line. By the result of Exercise (4.5.2), E(/) = E(J) = 2/n; hence
Z =I+ J satisfies E(~Z) = 2/n.
191
[4.5.4][ 4.5.6]
Solutions
in,
z i
~ jj
dz de
(z,ll):
O<z<i min(sinll,cosll}
0<11<i1f
1r
v2
1r
and hence
1
1
4
1
V.
3./2
4
var(,. Z) =    +  (2  v 2) =     .
..
7r
7r2
7r
7r
7r2
(iii) For Buffon's needle, the variance of the number of intersections is (2/n) (2/n )2 which exceeds
var(i Z). You should therefore use Buffon's cross.
JJ
fe;,
jj g(x)h(y)fx,y(x, y)dx dy
g(x)fx(x)dx
h(y)fy(y)dy = lE(g(X))lE(h(Y)).
(ii) By independence
6. If 0 is the centre of the circle, take the radius OA as origin of coordinates. That is, A = (1, 0),
B = (1, E>), C = (1, <1>), in polar coordinates, where we choose the labels in such a way that
0 ::5 e ::5 <1>. The pair e, <I> has joint density function j(e, cp) = (2n 2 ) 1 for 0 < e < cp < 2n.
The three angles of ABC are
of pairs (&, cp) such that 0 < e <
cp
i<I>.
192
Solutions
[4.5.7][4.6.3]
Then integrate f over this region to obtain the result. The shape of the region depends on whether or
not x < ~. The density function g of the largest angle is given by differentiation:
g(x) = {
'f 3::::
1
1
2
6(3x 1)
6(1 x)
if~::::x::::l.
X::::
lE{X(Xr X)} =
~JE ( L
XrXs) JE(X2 ) =
8.
+ JE(X) var(Y) =
0.
9. If X and Yare positive, then S positive entails T positive, which displays the dependence. Finally,
S2 =X and T 2 = Y.
1
f(B, </>)=I cos<f>l,
4rr
o:::: e <
2rr.
The marginals are then fe(B) = (2rr) 1 , fq,(<f>) = ~I cos </>I, and the conditional density functions
are
for appropriate e and</>. Thus E> and <I> are independent. The fact that the conditional density functions
are different from each other is sometimes referred to as 'Borel's paradox'.
2.
We have that
1/f(x) =
1oo Y !x,y(x, y) dy
oo fx(x)
and therefom
lE(l/f(X)g(X)) =
=
y(x y)
'
' g(x)fx(x)dxdy
1oooo 1oo y fxfx(x)
L: L:
~
3. Take Y to be a random variable with mean oo, say fy (y) = y 2 for 1 :::: y < oo, and let X = Y.
Then JE(Y 1 X) = X which is (almost surely) finite.
193
[4.6.4][4.6.6]
4.
Solutions
so that fYIX (y I x)
(b) Similarly,
so that fYIX(Y
5.
X< 00,
0 ::S
X< 00,
fx(x)
0 ::S
y < oo.
We have that
lP'(Y
= k) =
=
loo
lP'(Y
(n)
= k I X= x)fx(x) dx = fo1
o
+ k,
B(a
(n)
n
xk(l  x)nk
xa1(1 _ x)b1
B(a,b)
k +b).
B(a, b)
=b=
1, this yields
lP' Y =k)
(
(n)
k
= _1_
n + 1,
:::=:
k :::=: n,
loo
E(Y
na
I X= x)fx(x) dx =   ,
a+b
+ b +n)
var(Y) = (a+,b)""2,(a+_____,.b+l)
6.
By conditioning on X 1,
Gn(x) = lP'(N > n) =fox Gn1 (xu) du =fox Gn1 (v) dv.
EN=
n=O
More generally,
GN(s)
oo
oo
= L:snlP'(N
= n) = L:sn
n=1
whence var(N)
n=1
n) =
n1
x
 ::..._
(nl)!
n!
(s l)esx
+ 1,
dx
Solutions [4.6.7][4.7.1]
lEY
0. By the CauchySchwarz
Hence,
lE(var(Y
8.
I X)) = lE(Y 2 )
lE(lE(Y
I X) 2 )
lE(XY) 2
lE(X 2 )
< JEY 2 
(1  p 2 ) var(Y).
Another way is to observe that min{Y, Z} is exponentially distributed with parameter ft + v, whence
lP'(X < min{Y, Z}) = /...f(/... + ft + v). Similarly, lP'(Y < Z) = ~t/(1t + v), and the product of these
two terms is the required answer.
9.
gcy 3 eY,
fx(x)
00
= y) = 1Y and JE(Y I X = x) = x + 2.
I x),
10. We have that N > n if and only if Xo is largest of {Xo, X 1, ... , Xn}, an event having probability
lP'(XN :::; x)
F(x)n+l
n(n + 1)
n=1
oo
F(x)n+l
n

n=1
F(x)n+l
n+1
oo
= n}, Xn is the
+ F(x),
n=1
as required. Finally,
lP'(M
= m) = lP'(Xo :=:: X1
= m!
 (m
+ l)!
lP'(XY :::; u)
= lP'(XY
:::; u, Y :::; u)
+ lP'(XY :::; u, Y
> u)
= lP'(Y
:::; u)
+ lP'(X:::; ufY,
Y > u)
1u
=u+
dy=u(11ogu).
u y
lP'(XY:::; u,
z 2 :::; v) = lP'(XY
195
= u.,fV(l
log u),
0 < u, v < 1.
[4.7.2][4.7.5]
Solutions
g(u, v 
0:::: u, v::::
Hence
JJ
lP'(XY:::: Z 2 ) =
1.
~
lo;%u) dudv =
o::::u::::v::::1
Arguing more directly,
Iff
lP'(XY:::: Z 2 ) =
dxdydz =
o::::x,y,z::::1
xy::::z 2
2.
=I
vv
1
u
u
I=
u
Hence Ill= lui, and therefore fu. y(u, v) = ueu, for 0:::: u < oo, 0:::: v:::: 1. Hence U and V are
independent, and fv(v) = 1 on [0, 1] as required.
3.
Arguing directly,
2
0::::
so that fy (y) =
variables.
4.
(a) lP'(sin 1 X :::: y) = lP'(X :::: siny) = siny, for 0 :::: y ::::
o::::y::::in.
(b) Similarly, lP'(sin 1 X:::: y) =
_ln
< Y <
ln.
2  2
y:::: 1,
in.
I~ Vl~p 2 1 = R
= {w >
0,
> Wp
I R},
~JT
oo
1
I 2
ezr rdrdB =
Je=a r=O 2n
196
~JT 1
dB
2n
Solutions [4.7.6][4.7.9]
6.
U, Y
pU +
JI=P2 V, we have that U and V are independent N(O, 1) variables. It is easy to check that Y > X
if and only if (1  p) U < JI=P2 V. Turning to polar coordinates,
E(max{X, Y}) =
1 ~ [~1{1+rr
00
where tan 1{1' = ~(1 p)/(1 + p). Some algebra yields the result. For the second part,
E(X 2 ) +
We have that
lP'(X < Y, Z > z)
(a) lP'(X
= Z) = lP'(X <
Y)
A
= lP'(z <X< Y) = e(A+JL)z
= lP'(X
A+JL
= .
A+JL
(b) By conditioning on Y,
lP'( (X Y)+
w)
= _1L_eJ.w
A+JL
for w > 0.
By conditioning on X,
lP'(V > v)
= lP'(IX Yl
> v)
00
00
v > 0.
(c) By conditioning on X, the required probability is found to be
8.
that
Either make a change of variables and find the Jacobian, or argue directly. With the convention
r2  u2 = 0 when r 2  u 2 < 0, we have that
~
7r
a2 F
2r
f(r,x) =  =
~
arax
rry r2 x2
9.
jx
Jr2 u2du,
r
4n r
197
[4.7.10][4.7.14]
Solutions
= ~r for lzl < r < 1. This question may be solved in spherical polars also.
The transformations = x + y, r = xf(x + y), has inverse x = rs, y = (1  r)s and Jacobian
Hence f(r, z)
10.
J = s. Therefore,
fR(r)
lo
lo
r)s )s ds
O::=or::=ol.
whence
fy(y) = 2fx(J(afy)
1)
1
rrJy(a y)
0 :::0 y :::0 a.
12. Using the result of Example (4.6.7), and integrating by parts, we obtain
JP(X >a, Y >b)=
00
= [1 
<f>(x) { 1 <I> (
~)} dx
<l>(a)][1  <l>(c)]
00
[1  <l>(x)]</> (
~)
1 p2
1 p2
dx.
Since [1  <l>(x)]/<f>(x) is decreasing, the last term on the right is no greater than
1  <l>(a)
</>(a)
00
<f>(x)<f>
b px )
JT=P2 JT=P2
dx,
P(Y > a)
by the transformation v
= JP(O <
1
X < a
ra1 rr(l +
du
= Jo
u2)
1aoo rr(1 +
v2 dv
v2),
f(x) = { 2
'
ifO:Sx:Sl.
A.(A.w)a+.BleA.w
r(a
+ {3)
za1(1 z).Bl
B(a, {3)
Hence Wand Z are independent, and Z is beta distributed with parameters a and {3.
198
Solutions [4.8.1][4.8.4]
rz AJ.LeAxeJL(zx) dx = ~
(eAZ eJLZ),
11 A
z ~ 0,
lo
if A ::j:. fL. What happens if A = J1? (Z has a gamma distribution in this case.)
2.
Using the convolution formula (4.8.2), W =aX+ f3Y has density function
fw(w)
00
oo
na(1 + (x/a) 2)
1
nf3(1 + {(w x)/f3}2)
dx,
1 .
1
af3 .  dz
z2 + a2 (z w)2 + p2
JD n2
where D is the semicircle in the upper complex plane with diameter [ R, R] on the real axis. Evaluating the residues at z = ia and z = w + if3 yields
af32ni { 1
1
1
1
}
fw(w) = _n_2_ 2ia (ia w)2 + p2 + 2if3 (w_+_i_f3,)2,+a"""
2
< w <
00
00
4.
~zez dy = ~z 2 ez,
z ~ 0.
_ , J.. 1x
f 2 (x ) II.Je
A
2
A2  At
A
2
n
A
+ "2e
, J..2x  1
_ '"""" , J..,x II
  L...J"re
 s .
At  A2
r=l
s=l As  Ar
sf.r
fn(X) = LAreJ..,x
r=l
IIn __As_,
s=l As  Ar
sf.r
which may be proved by induction as follows. Assume that (*) holds for n ::::: N. Then
199
[4.8.5][4.8.8] Solutions
1=L IT _ s _ +A ,
N N+1
r=1 s=1 As  Ar
sf.r
5.
=N
AN+1
+ 1 on solving for A.
fz(x) = {
ifO:::::x:::::1,
2
if 1 ::::::
X ::::::
2.
f3(x)=
lo1fz(xy)dy= 11
0
(xy)dy+
lox1 (2x+y)dy=~(x~) .
2
x1
Likewise,
= 0, as required.
7. From the representation X= apU + a.Jl=PZV, Y = r:U, where U and V are independent
N(O, 1), we learn that
apy
E(X I Y = y) = E(apU I U = yfr:) =  .
r:
Similarly,
whence var(X 1 Y) = a 2 (1  p 2 ). For parts (c) and (d), simply calculate that cov(X, X + Y)
a 2 +par:, var(X + Y) = a 2 + 2par: + r: 2 , and
r:2(1  p2)
1 p(X, X+ Y) =
+ 2par: + r: 2
First recall that lP'(IXI :::::: y) = 2<1>(y)  1. We shall use the fact that U = (X+ Y)j,fi,
V = (X  Y) I ,J2 are independent and N (0, 1) distributed. Let 8.. be the triangle of JR 2 with vertices
(0, 0), (0, Z), (Z, 0). Then
8.
8.)
= 2{2<1>(z!..fi) 1}2 ,
200
by symmetry
Solutions [4.9.1][4.9.5]
f(z)
2J2{2<t>(z!J2) l}tP(z;J2).
Finally,
lE(Z
IX
IX
> 0, Y > 0)
I X>
0) = 41E(X/[X>O}) = 4
loooo
I 2
r;c:e2x dx.
v2rr
as required. Clearly W is nonsingular if and only if A 2 is nonsingular. This happens if and only if
Ai > 0 for all i, which is to say that V is positive definite.
By Theorem (4.9.6), Y has the multivariate normal distribution with mean vector 0 and covariance
matrix
2.
Clearly Y = (X  J.L)a' + J.La' where a = (a1, az, ... , an). Using Theorem (4.9.6) as in the
previous solution, (X  J.L )a' is univariate normal with mean 0 and variance aVa'. Hence Y is normal
with mean J.La' and variance aVa'.
3.
!.
,2.(x2 2pxy
2(1  p )
+ Y, V
and therefore U = X
+ i) = 
2 { u2(1 p)
4(1 p )
+ v), y
= !<u v), so
+ v2(1 + p) },
1
{
u2
v2
}
f(u, v) = 4rrv'l=P2 exp  4(1 + p)  4(1 p) ,
whence U and V are independent with respective distributions N(O, 2(1
5.
y) = lP'(X
p )).
~
a,
a x 2 ifJ(x)dx
a
~a x 2 ifJ(x)dx
oo
201
[4.9.6][4.10.1]
Solutions
The answer to the final part is no; X and Y are N (0, 1) variables, but the pair (X, Y) is not bivariate
normal. One way of seeing this is as follows. There exists a root a of the equation p(a) = 0. With
this value of a, if the pair X, Y is bivariate normal, then X andY are independent. This conclusion is
manifestly false: in particular, we have that lP'(X > a, Y > a) # lP'(X > a)lP'(Y > a).
6.
Recall from Exercise (4.8.7) that for any pair of centred normal random variables
E(X I Y) = cov(X, Y) Y,
varY
var(X
I Y)
7. As in the above exercise, we calculate a = E(X I I ~7 Xr) and b = var(X I I ~7 Xr) using the
facts that var XI = vu, var(~7 Xi) = ~ij ViJ and cov(XI, ~7 Xr) = ~r VIr
8.
Let p = lP'(X > 0, Y > 0, Z > 0) = lP'(X < 0, Y < 0, Z < 0). Then
1 p = lP'({X > 0} U {Y > 0} U {Z > 0})
= lP'(X > 0)
+ lP'(Y
> 0)
+ lP'(Z
> 0)
+p
9.
[3
. I
. I
. I }]
2 + p  4 + 2rr1 {sm
PI + sm P2 + sm P3 .
J1 p[V,
z=
p 3U
2 P32 + 2PIP2P3
1 PI2  P2
+ P2 PIP3 V +
J1p[
(1 PI)
w.
2:0.
() lo x2
z lmI e !x(
._ z x )lnI
._
e l(zx)d
2
x
g z = c
= cz2(m+n Ie2z
I
I
Jo{I u'ZmI(lu)ZnI du
202
Solutions [4.10.2][4.10.6]
by the substitution u = xjz, where cis a constant. Hence g(z) = c'z~(m+n) 1 e~z for z.:,:: 0, for
an appropriate constant c', as required.
Second method. If m and n are integral, the following argument is neat. Let Z1, Z2, ... , Zm+n be
independent N(O, 1) variables. Then X1 has the same distribution as Zf + Zi + + Z~, and X2
the same distribution as Z~+ 1 + Z~+ 2 + + Z~+n (see Problem (4.14.12)). Hence X1 + X2 has
the same distribution as Zf + + Z~+n i.e., the x2(m + n) distribution.
2. (i) The t(r) distribution is symmetric with finite mean, and hence this mean is 0.
(ii) Here is one way. Let U and V be independent x2 (r) and x2 (s) variables (respectively). Then
E(V
1
lo0012s/2 's1'v
r(is1)lo002:z(s2) 's2'v
1
)=
v:Z
e :Z dv=
v:Z
e 2 dv=0
v r(is)
2f(is)
0
r(is 1)
s 2
s
( Ujr)
Vjs =s2
if s > 2.
3.
First method. Find the density function of X I Y, using a change of variables. The answer is
F(2, 2).
4.
x2 (2)
5. The vector (X, X1 X, X2 X, ... , Xn X) has, by Theorem (4.9.6), a multivariate normal
distribution. We have as in Exercise (4.5.7) that cov(X, Xr X) = 0 for all r, which implies that X
is independent of each X r. Using the form of the multivariate normal density function, it follows that
X is independent of the family {Xr X : 1 ::; r ::; n}, and hence of any function of these variables.
Now s2 = (n 1) 1 L:r(Xr X) 2 is such a function.
6. The choice of fixed vector is immaterial, since the joint distribution of the Xj is spherically
symmetric, and we therefore take this vector to be (0, 0, ... , 0, 1). We make the change of variables
U2 = Q2 +X~, tan W = Q/ Xn, where Q2 = 2::~;;;;;}
and Q .:,:: 0. Since Q has the x2(n  1)
distribution, and is independent of Xn, the pair Q, Xn has joint density function
x;
~x 2 l(lq)!(n3)e~x
/( x  _e__ . 2 2
q' )  ../27i ~r'(;i(n1)) 
X E
JR, q > 0.
The theory is now slightly easier than the practice. We solve for U, W, find the Jacobian, and deduce
the joint density function fu,'l!(u, 1/1) of U, W. We now integrate over u, and choose the constant so
that the total integral is 1.
203
[4.11.1][4.11. 7]
Solutions
2.
3. We may assume without loss of generality thatA = 1 (since ZIA. is f(A., t) if Z is f(l, t)). Let U,
V be independent random variables which are uniformly distributed on [0, 1]. We set X= t log V
and note that X has the exponential distribution with parameter 1It. It is easy to check that
for x > 0,
where c = tt et+ 1I f(t). Also, conditional on the event A that
U <

xt1 t
e
f(t)
teXft
X has the required gamma distribution. This observation may be used as a basis for sampling using
the rejection method. We note that A= {log U :=:: (n l)(log(XIn) (X In)+ 1) }. We have that
IF'( A) = 11c, and therefore there is a mean number c of attempts before a sample of size 1 is obtained.
4. Use your answer to Exercise (4.11.3) to sample X from f(l, a) andY from f(l, {3). By Exercise
(4.7.14), Z = XI(X + Y) has the required distribution.
S. (a) This is the beta distribution with parameters 2, 2. Use the result of Exercise (4).
(b) The required r (1, 2) variables may be more easily obtained and used by forming X =  log( U1U2)
and Y log(U3 U4) where {Ui : 1 _::: i :=:: 4} are independent and uniform on [0, 1].
(c) Let U1, U2, U3 be as in (b) above, and let Z be the second order statistic U(2) That is, Z is the
middle of the three values taken by the Ui; see Problem (4.14.21). The random variable Z has the
required distribution.
(d) As a slight variant, take Z = max{U1, U2} conditional on the event {Z :=:: U3}.
(e) Finally, let X = .../UlI (.../Ul + ..JU2), Y = .../Ul + ..jUi. The distribution of X, conditional on
the event {Y :=:: 1}, is as required.
6. We use induction. The result is obvious when n = 2. Let n 2: 3 and let p = (P1, P2, ... , Pn) be
a probability vector. Since p sums to 1, its minimum entry P(1) and maximum entry P(n) must satisfy
1
1
P(1) :S  <   ,
n
n 1
1
n 2
P =  1 v1 +  1 Pn1
nn
where
1P(1)
n 1
1+(n2)P(1)
1
2: 1
n 1
n .
n1
(
1
)
Pn1 =  0, P1 + P2   , P3, , Pn ,
n2
n1
is a probability vector with at most n  1 nonzero entries. The induction step is complete.
It is a consequence that sampling from a discrete distribution may be achieved by sampling from
a collection of Bernoulli random variables.
7. It is an elementary exercise to show that lP'( R2 :=:: 1) = TC, and that, conditional on this event, the
vector (T1, T2) is uniformly distributed on the unit disk. Assume henceforth that R 2 :=:: 1, and write
(R, 8) for the point (T1, T2) expressed in polar coordinates. We have that R and e are independent
with joint density function fR,e(r, 8) = r In, 0 :=:: r :=:: 1, 0 :=:: () < 2n. Let (Q, W) be the polar
204
Solutions
e and e2I Q2
[4.11.8][4.12.1]
I 2
independent, and, by a change of variables, Q has density function fQ(q) = qe1q , q > 0. We
recognize the distribution of (Q, 111) as that of the polar coordinates of (X, Y) where X and Y are
independent N(O, 1) variables. [Alternatively, the last step may be achieved by a twodimensional
change of variables.]
8.
We have that
9.
2r
,
T(
JP(Z
IT
(1 h(r))
r=O
JP(X > O)JP(X > 1 I X> 0) .. li"(X
= m) = h(m)
= m I X>
m 1)
= JP(X = m).
11. Suppose g is increasing, so that h(x) =  g(1 x) is increasing also. By the FKG inequality of
Problem (3.11.18b), K = cov(g(U),g(1  U)) 2::: 0, yielding the result.
Estimating I by the average (2n)  1 I:~~ 1 g (Ur) of 2n random vectors U r requires a sample
of size 2n and yields an estimate having some variance 2na 2 . If we estimate I by the average
(2n) 1 {2:~= 1 g(Ur) + g(1  Ur) }, we require a sample of size only n, and we obtain an estimate
with the smaller variance 2n(a 2  K).
12. (a) By the law of the unconscious statistician,
lE [g(Y)fx(Y)J
fy(Y)
g(y)fx(Y)f ( )d =I.
fy(y)
y y y
(b) This is immediate from the fact that the variance of a sum of independent variables is the sum of
their variances; see Theorem (3.3.11b).
(c) This is an application of the strong law of large numbers, Theorem (7.5.1).
13. (a) If U is uniform on [0, 1], then X = sin(~rr U) has the required distribution. This is an example
of the inverse transform method.
(b) If U is uniform on [0, 1], then 1 U 2 has density function g(x) = {2~}  1 , 0 ::=: x ::=: 1.
Now g(x) 2::: (rr/4)/(x), which fact may be used as a basis for the rejection method.
Suppose that lE(u(X)) 2::: lE(u(Y)) for all increasing functions u. Let c E ffi. and set u
lc(x) = {
1 if X >
.
0 If X :S
205
C,
C,
= Ic where
[4.12.2][4.13.1]
Solutions
4.
For any A
R,
lP'(X
#
#
1\
y = y, where x
fx(k)
1\
y = min{x, y}. It
fy(k),
and by the definition of dTv(X, Y) that the common value in this display equals ~dTv(X, Y) = 8.
Let U be a Bernoulli variable with parameter 18, and let V, W, Z be independent integervalued
variables with
IP'(V = k) = Ux(k) fy(k)}+ /8,
lP'(W
= k) = {fy(k)
IP'(Z = k) = fx(k)
 fx(k)}+ /8,
A
fy(k)/(1  8).
Then X'= UZ + (1 U)V andY'= UZ + (1 U)W have the required marginals, and IP'(X' =
Y') = lP'(U = 1) = 1  8. See also Problem (7.11.16d).
6. Evidently dTv(X, Y) = Jp qJ, and we may assume without loss of generality that p 2: q. We
have from Exercise (4.12.4) that lP'(X = Y) _::: 1  (p q). Let U and Z be independent Bernoulli
variables with respective parameters 1 p+q andq/(1p+q). The pair X'= U(Z1)+ 1, Y' = UZ
has the same marginal distributions as the pair X, Y, and IP'(X' = Y') = IP'(U = 1) = 1  p + q as
required.
To achieve the minimum, we set X" = 1  X' and Y" = Y', so that IP'(X" = Y") = 1 IP'(X' =
Y') = P q.
in+
206
Solutions [4.13.2][4.13.6]
Geometrical probability
2. Let A be the left shaded region and B the right shaded region in the figure. Writing A for the
random line, by Example (4.13.2),
IP'(A meets both Stand S2)
3. With III the length of the intercept I of Al with S2, we have that IP'(A2 meets I)
by the Buffon needle calculation (4.13.2). The required probability is
1 [ 2n
2 Jo
4.
21II
00
dpdB
[ 2n
= Jo
IS2I
b(S 1) 2 dB
= 21II/b(St),
2niS21
b(S 1) 2 .
If the two points aredenotedP = (Xt, Yt), and Q = (X2, Y2), then
We use Crofton's method in order to calculate lE(Z). Consider a disc D of radius x surrounded by an
annulus A of width h. We set A(x) = lE(Z I P, Q E D), and find that
D, Q
A)
=
nx
k~nk2xcos0 2
r dr dB
o o
c:
+ o(1)
+ o(h)) .
32x
= ,
9n
whence
dA
4A
128
dx
x
9n
which is easily integrated subject to A(O) = 0 to give the result.
=+,
5. (i) We may assume without loss of generality that the sphere has radius 1. The length X = IAOI
has density function f(x) = 3x 2 for 0 :::=: x :::=: 1. The triangle includes an obtuse angle ifB lies either
in the hemisphere opposite to A, or in the sphere with centre ~X and radius ~X, or in the segment
cut off by the plane through A perpendicular to AO. Hence,
(ii) In the case of the circle, X has density function 2x for 0 :::=: x :::=: 1, and similar calculations yield
1
1
1
~
3
IP'(obtuse) =  + + lE(cos 1 X Xy 1 X2) = .
6.
JT
(X, Y) and G
= (Yt, Y2),
= ~IABilE(Y) = ~IABin =
207
IABGI.
[4.13.7][4.13.11]
7.
Solutions
With b = lAB I and h the height of the triangle ABC on the base AB, we have that IGt G2l = ~ b and
the height of the triangle AGt G2 is ~h. Hence,
lEIAPQI = ~ ~b ~h = ~IABq.
8. Let the scale factor for the random triangle be X, where X E (0, 1). For a triangle with scale
factor x, any given vertex can lie anywhere in a certain triangle having area (1  x) 2 1ABq. Picking
one at random from all possible such triangles amounts to supposing that X has density function
f(x) = 3(1  x) 2 , 0::::: x ::::: 1. Hence the mean area is
9.
z ::::: a,
a )
F(z, a+ da) = F(z, a) (  
a+da
= F(z, a)
( 1  2da)
a
+JP>(X :::::_a z)
2ada
2 +o(da)
(a+da)
+az 2da
+ o(da),
a
and the equation follows by taking the limit as da .j, 0. The boundary condition may be taken to be
F(a, a) = 1, and we deduce that
2z F(z, a) = ;;
(z;; )2 ,
0 ::::: z ::::: a.
2da)
2da
mr(a+da)=mr(a) ( 1~ +E((axn~+o(da).
Now, JE((aXY) = ar /(r+ 1), yieldingtherequiredequation. The boundaryconditionismr (0) = 0,
and therefore
2ar
mr(a)
(r
+ 1)(r + 2)
10. If n great circles meet each other, not more than two at any given point, then there are 2G)
intersections. It follows that there are 4(~) segments between vertices, and Euler's formula gives the
number of regions as n(n  1) + 2. We may think of the plane as obtained by taking the limit as
R + oo and 'stretching out' the sphere. Each segment is a side of two polygons, so the average
number of sides satisfies
4n(n 1)
+4
2 +n(n 1)
as n+ oo.
11. By making an affine transformation, we may without loss of generality assume the triangle has
vertices A= (0, 1), B = (0, 0), C = (1, 0). With P = (X, Y), we have that
L=
c~
y,
0) ,
M=
(X: y,
208
X : y) ,
N= (
0, 1~X) .
Solutions [4.13.12][4.14.1]
Problems
Hence,
lEJBLNJ = 2
ABC
xy
dxdy =
2(1x)(l y)
in
2 
klo (
x
) dx =11:2
3
x logx
1x
6
2'
= 41P'(S
PQR)
= 4 lz.
13. We use Crofton's method. Let m(a) be the mean area, and condition on whether points do or do
not fall in the annulus with internal and external radii a, a+ h. Then
m(a+h)=m(a)C:hr
+[~ +o(h)Jm(a).
where m(a) is the mean area of a triangle having one vertex P on the boundary of the circle. Using
polar coordinates with P as origin,
dm
da
6m
a
6
a
35a 2
36n
=+,
whence m(a) = (35a 2 )1(48n).
14. Let a be the radius of C, and let R be the distance of A from the centre. Conditional on R, the
required probability is (a  R) 2 I a 2 , whence the answer is JE((a  R) 2 I a 2 ) =
(1  r )2 2r dr =
i.
JJ
15. Let a be the radius of C, and let R be the distance of A from the centre. As in Exercise (4.13.14),
the answer is lE((a R) 3 la 3 ) =
(1 r) 3 3r 2 dr = 2~.
JJ
(x 
f.L)I(a./2).
I
I 2
I 2
(b) The mean is J~00 x (2n) 'Z e 'Z x dx, which equals 0 since x e 'Z x is an odd integrable function.
I
I 2
The variance is J~00 x 2 (2n) 'Z e 'Z x dx, easily integrated by parts to obtain 1.
209
[4.14.2][4.14.5]
Solutions
I 2
and also 1  3y 4 < 1 < 1 + y 2. Multiply throughout these inequalities by ezY f./iii, and
integrate over [x, oo ), to obtain the required inequalities. More extensive inequalities may be found
in Exercise (4.4.8).
(d) The required probability is a(x) = [1  <t>(x + ajx)]/[1  <t>(x)]. By (c),
as x+ oo.
2.
Clearly
<
f3 : :=: 1. Also
3. The Ai partition the sample space, and i  1 : :=: X(w) < i if wE Ai Taking expectations and
using the fact that JE(/i) = JP>(Ai), we find that S::::: lE(X) ::::: 1 + S where
00
s=
00
i=2
4.
i1
00
00
00
j=1 i=j+1
j=1
= JP>(F(X)
= PR,
:::: eY)
= 1
0::::: y:::::
1.
eY if y :::: 0.
+tan 1 d).
Clearly
JP>(X > s + x
IX
> s) =
J!D(X > s + x)
e}.(s+x)
JP>(X > s) =
e}.s
= eAx
210
Solutions [4.14.6][4.14.9]
Problems
rationals u and v satisfying v ::0 x ::0 u. Hence g(l)" ::0 g(x) ::0 g(l)v. Take the limits as u {. x and
v t x through the rationals to obtain g (x) = eJLx where 1i = log g ( 1).
= g(x)
and
1=
Hence fx(x)fy(y)
r:
r:
= g(x)h(y) =
h(y) dy,
fy(y)dy
fx and h
fy(y)
= h(y)
J: f:
g(x)dx
r:
g(x) dx
h(y)dy.
fx(x)
=1 2exy dy =2e x,
00
=2eY(l eY),
Jx=O }y=x
8.
As in Example (4.13.1), the desired property holds if and only if the length X of the chord satisfies
X ::0 ,J3. Writing R for the distance from P to the centre 0, and E> for the acute angle between the chord
sin E>, and therefore J!D(X ::0 ,J3) = JP>(R sinE> ::::: ~ ).
and the line OP, we have that X = 2 1 The answer is therefore
lP'
R2 2
(R > _1_)
= ~ r!rr lP' (R > _1_) d(}
2sinE>
rr Jo
 2sin8
'
9.
Evidently,
E(U)
Secondly,
= E(V 2 )  ! Jo
211
dx ::0 E(V 2 ).
[4.14.10][4.14.11]
Hence var(W)
:::=::
Solutions
var(V)
:::=::
var(U).
10. Clearly the claim is true for n = 1, since the r(A., 1) distribution is the exponential distribution.
Suppose it is true for n :::=:: k where k :=::: 1, and consider the case n = k + 1. Writing fn for the density
function of Sn, we have by the convolution formula (4.8.2) that
+ 1) density function.
11. (a) Let Z1, Z2, ... , Zm+n be independent exponential variables with parameter A.. Then, by
Problem (4.14.10), X' = Z1 + + Zm is f(A, m), Y' = Zm+1 + + Zm+n is r(A., n), and
X'+ Y' is f(A., m + n). The pair (X, Y) has the same joint distribution as the pair (X', Y 1), and
therefore X+ Y has the same distribution as X'+ Y', i.e., f(A., m + n).
(b) Using the transformation u = x + y, v = xf(x + y), with inverse x = uv, y = u(l  v), and
Jacobian
I
J we find that U
fu
v v
1
1
u  u '
u
'
v(u, v)
v)n1}
for u :=::: 0, 0 :::=:: v :::=:: 1. Hence U and V are independent, U being f(A., m + n), and V having the beta
distribution with parameters m and n.
(c) Integrating by parts,
m1
(A.tY
k!
k=O
(d) This may be achieved by the usual change of variables technique. Alternatively, reflect that, using
the notation and result of part (b), the invertible mapping u = x + y, v = xf(x + y) maps a pair
X, Y of independent (r(A., m) and f(A., n)) variables to a pair U, V of independent (f(A., m +n) and
B(m, n)) variables. Now UV =X, so that (figuratively)
'T(A., m
+ n) x
212
Solutions [4.14.12][4.14.15]
Problems
12. (a) Z
= xf satisfies
z 2:0,
= xf +X~ satisfies
z 2:0,
for some constant c. This is the x2 (n
f XIY (x I Y ) =
2
Jx,y(x,y) = c1 ()
{
1
(x
2xp,l
y exp     2px(yM))}
fy(y)
2(1  p 2 ) a[
a[
awz
X E
IR,
for some cz(y). This is the normal density function with mean 111 +pal (y p,z)/az and variance
a[O  p 2 ). See also Exercise (4.8.7).
14. Set u = yjx, v = x, with inverse x = v, y = uv, and Ill= lvl. Hence the pair U = Y/X,
V = X has joint density fu. v(u, v) = fx.r(v, uv)lvl for oo < u, v < oo. Therefore fu(u) =
f~oo f(v, uv)lvl dv.
15. By the result of Problem (4.14.14), U = Y I X has density function
fu(u)
L:
f(y)f(uy)IYI dy,
and therefore it suffices to show that U has the Cauchy distribution if and only if Z = tan l U is
uniform on (~n, ~n). Clearly
IP'(Z::; 0) = IP'(U :StanO),
213
~n <
e<
~n,
[4.14.16][4.14.17]
Solutions
= fu(tanO)sec 2 e.
whence fz(O)
Therefore fz(O)
= n 1 (for 101
fu(u) = n(l
When
+ u2)'
00
< j;n)ifandonly if
1oooo
f(x)f(xy)lxldx = 2
looo e2x
1
(l+y )xdx,
1 2
o 2n
which is easily integrated directly to obtain the Cauchy density. In the second case, we have the
following integral:
1oo
a2 1xl
00
0:::::
e < 2n.
Therefore Rand 8 are independent, 8 being uniform on [0, 2n), and R 2 having distribution function
I 2
= re2r
x2 (2)).
The
for r > 0.
In the first octant, i.e., in {(x, y) : 0 :::=:: y :::=:: x}, we have min{x, y} = y, max{x, y} = x. The
joint density Jx,Y is invariant under rotations, and hence the expectation in question is
!o
y
Jx,y(x,y)dxdy=8
O.:::;y.:::;x X
lerr:f41oo
1:1=0
r=O
tane
1 2
2
 2re'Zr drd0=log2.
7i
7i
:::=::
Similarly
lP'(V
:::=::
u)
:::=::
v)
= 1 e 2u for u :=::: 0.
+ j; Y satisfies
lP'(Z > v) =
= e2v(ev
1)
v).
00
ex dx
Solutions [4.14.18][4.14.19]
Problems
Therefore lE(V) = lE(X)
+ ilE(Y) =
+ ! var(Y) =
i by independence.
w, X > Y).
Now
JP>(U :S u, W > w, X :S Y) = J!D(X::: u, Y > X+ w) =lou .l..e}..xeJL(x+w) dx
= _J.._eJLW(l e(A+JL)U)
.l..+tt
and similarly
+w <
oo,
an expression which factorizes into the product of a function of u with a function of w. Hence U and
W are independent.
19. U =X+ Y, V =X have joint density function fy(u v)fx(v), 0::: v::: u. Hence
fv u(v I u) = !u, v(u, v) =
1
fu(u)
(a) We are given that fvJU(V
I u)
u fy(u v)fx(v)
.
fy(u y)fx(y)dy
fo
fy(O)fx(u)
by setting v = 0
by setting v = u.
In particular fy(u) and fx(u) differ only by a multiplicative constant; they are both density functions,
implying that this constant is 1, and fx = fy. Substituting this throughout the above display, we
find that g(x) = fx(x)ffx(O) satisfies g(O) = 1, g is continuous, and g(u v)g(v) = g(u) for
0::: v ::: u. From the hint, g(x) = e}..x for x 2:: 0 for some./.. > 0 (remember that g is integrable).
(b) Arguing similarly, we find that
ru
c
fy(u v)fx(v) = ua+.B 1 va 1(u v).B 1 Jo fy(u y)fx(y)dy
215
[4.14.20][4.14.22]
Solutions
for 0::: v::: u and some constant c. Setting fx(v) = x(v)va 1 , fy(y) = 1J(y)yf3 1, we obtain
IJ(U v)x (v) = h(u) for 0::: v::: u, and for some function h. Arguing as before, we find that 17 and
x are proportional to negative exponential functions, so that X and Y have gamma distributions.
20. We are given that U is uniform on [0, 1], so that 0 ::: X, Y :::
<
i.
and similarly
E = J!D(X + Y > 1 E) :S J!D(X >
! E) ::':
JE.
E) 2 ,
Now
2.
Therefore all the above inequalities are in fact equalities, implying that J!D(X < E) = J!D(X >
JE if 0 <
<
i. Hence a contradiction:
1 E) =
1
1 X+ Y ::': g)<
1
1
1
g1 = JP>(X + Y < g)=
J!D(X, Y < g1) J!D(X, Y < g
J!D(X < 81 , Y < g)=
s
21. Evidently
LlP' (Xrr
Jf
where the sum is over all permutations n = (n1, nz, ... , l!n) of (1, 2, ... , n). By symmetry, each
term in the sum is equal, whence the sum equals
n!J!D(X1 :S Y1 , Xn :S Yn. X1 < Xz < < Xn).
The integral form is then immediate. The joint density function is, by its definition, the integrand.
22. (a) In the notation of Problem (4.14.21), the joint density function of X(2), ... , X(n) is
gz(yz, , Yn)
Y2
00
nl
k1
f(Yk) f(Yn).
216
Solutions [4.14.23][4.14.25]
Problems
(b) It is neater to argue directly. Fix x, and let I r be the indicator function of the event {Xr ::::; x}, and
letS= It + h ++ln. Then Sis distributed as bin(n, F(x)), and
lP'(X(k) ::::; x) = lP'(S;:: k) =
t (;)
l=k
t (;)
l=k
1
kn(nl)(nk+l) k 1 (
x)nk
 fk(x jn) = x
1 n
k!
nk
n
as n
1
k x
   x 1e
(k  1)!
(b) For an increasing sequencex(l), X(2) ... , X(n) in [0, 1], we define the sequence Un = n 1ogx(n)
Uk = klog(x(k)/X(k+l)) for 1 ::::; k < n. This mapping has inverse
X(k) = X(k+1)e
u
kl
n
= exp {  ~
1
1
u;
l=k
with Jacobian J = ( l)neulu2un jn!. Applying the mapping to the sequence X(l) X(2) ... ,
X(n) we obtain a family U1, U2, ... , Un of random variables with joint density g(u1, u2, ... , un) =
eulu2un for u; ;:: 0, 1 ::::; i ::::; n, yielding that the U; are independent and exponentially
distributed, with parameter 1. Finally log X (k) =  'L/t=k i  1U;.
(c) In the notation of part (b), Zk = exp( Uk) for 1 ::::; k ::::; n, a collection of independent variables.
Finally, Uk is exponential with parameter 1, and therefore
lP'(Zk ::::; z) = lP'(Uk :=:: log z) = elogz = z,
O<z:Sl.
25. (i) (X 1, X2, X 3) is uniformly distributed over the unit cube of R 3, and the answeris therefore the
volume of that set of points (x1, x2, x3) of the cube which allow a triangle to be formed. A triangle
is impossible if x1 ;:: x2 + x3, or x2 ;:: x1 + x3, or x3 ;:: x1 + x2. This defines three regions of
the cube which overlap only at the origin. Each of these regions is a tetrahedron; for example, the
region x3 :=:: x1 + x2 is an isosceles tetrahedron with vertices (0, 0, 0), (1, 0, 1), (0, 1, 1), (0, 0, 1),
with volume ~. Hence the required probability is 1  3 ~ = ~.
(ii) The rods oflength x1, x2, ... , Xn fail to form a polygon if either x1 ;:: x2 + x3 + + Xn or any of
the other n  1 corresponding inequalities hold. We therefore require the volume of thendimensional
hypercube with n comers removed. The inequality x1 ;:: x2 + x3 + + Xn corresponds to the convex
hull of the points (0, 0, ... , 0), (1, 0, ... , 0), (1, 1, 0, ... , 0), (1, 0, 1, 0, ... , 0), ... , (1, 0, ... , 0, 1).
217
[4.14.26][4.14.27]
Solutions
Mapping x1 1+ 1  x1, we see that this has the same volume Vn as has the convex hull of the origin
0 together with then unit vectors e1, e2, ... , en. Clearly V2 = ~.and we claim that Vn = 1/n!.
Suppose this holds for n < k, and consider the case n = k. Then
where Vk1 (0, x1 e2, ... , x1 ek) is the (k I)dimensional volume of the convex hull ofO, x1 e2, ... ,
x1ek. Now
so that
{1
xk1
vk = lo
(k
~ l)! dx1 =
1
k!.
1X2<~,
or:
1
x1
< ~
t,
Plot the corresponding region of JR 2. One then sees that the area of the region is which is therefore
the probability in question.
(ii) The pieces may form a polygon if no piece is as long as the sum of the lengths of the others. Since
the total length is 1, this requires that each piece has length less than ~. Neglecting certain null events,
this fails to occur if and only if the disjoint union of events Ao U A 1 U U An occurs, where
Ao = {no break in (0,
~J},
remember that there is a permanent break at I. Now lP'(Ao) = (~)n, and fork 2: 1,
_....:.1X__:._l......{EIXPI}1/p'
IYiq
+ qEIYql
IXYI
2: {EIXPI}l/P{EIYql}l/q.
218
Solutions [4.14.28][4.14.31]
Problems
(b) We have, with Z =IX+
Yl.
xfxiY z(x I y, z) d x =
'
f.
Then
f(x,y,z)d
x.
!r,z(y, z)
Hence
E{E(X I Y, Z) Y =
y} =
=
rr
11 fr.z(y, z) fy(y)
f(x,y,z)
~r x fy(y) dxdz=E(XIY=y).
Alternatively, use the general properties of conditional expectation as laid down in Section 4.6.
30. The first car to arrive in a car park of length x + 1 effectively divides it into two disjoint parks
of length Y and x  Y, where Y is the position of the car's leftmost point. Now Y is uniform on
[0, x], and the formula follows by conditioning on Y. Laplace transforms are the key to exploring the
asymptotic behaviour of m (x) / x as x + oo.
31. (a) If the needle were oflength d, the answer would be 2/:rr as before. Think about the new needle
as being obtained from a needle of length d by dividing it into two parts, an 'active' part of length L,
and a 'passive' part of length d L, and then counting only intersections involving the active part.
The chance of an 'active intersection' is now (2/:rr)(L/d) = 2Lj(:rrd).
(b) As in part (a), the angle between the line and the needle is independent of the distance between
the line and the needle's centre, each having the same distribution as before. The answer is therefore
unchanged.
(c) The following argument lacks a little rigour, which may be supplied as a consequence of the
statement that S has finite length. ForE > 0, let XI, x2, ... , Xn be points on S, taken in order along
S, such that xo and Xn are the endpoints of S, and Ixi+ 1  Xi I < E for 0 ::::; i < n; lx  y I denotes
the Euclidean distance from x to y. Let Ji be the straight line segment joining Xi to xi+ 1, and let /i
be the indicator function of {Ji n A :1= 0}. If E is sufficiently small, the total number of intersections
between Jo U lt U U ln1 and S has mean
n1
'L:EUi)
i=O
2 n1
=lxi+l xil
:rrd i=O
219
[4.14.32][4.14.34]
Solutions
by part (b). In the limit as E .j.. 0, we have that L::i JE(/i) approaches the required mean, while
2L(S)
nI
:rrd L
lxi+l xil+   .
i=O
:rrd
32. (i) Fix Cartesian axes within the gut in question. Taking one end of the needle as the origin,
the other end is uniformly distributed on the unit sphere of R 3 . With the Xray plate parallel to the
(x, y )plane, the projected length V of the needle satisfies V ~ v if and only if IZ I ::::; ~. where
Z is the (random) zcoordinate of the 'loose' end of the needle. Hence, for 0::::; v ::::; 1,
;::;:)
;::;:)
4:rr ~
;::;:)
lP'(V~v)=lP'(Ylv~::;z::;vlv2)=
=vlv 2 ,
4:rr
since 4:rr ~is the surface area of that part of the unit sphere satisfying lzl ::::; ~(use
Archimedes's theorem of the circumscribing cylinder, or calculus). Therefore V has density function
e,
w =tan ~MA2
2.
1 B
Some slog is now required to calculate the Jacobian J of this mapping, and the answer will be
!A,B(a, b)= 41ll:rr 2 for 0 <a, b < 1, a 2 + b 2 > 1.
on the set I of increasing sequences of positive reals. Define the oneone mapping from I onto
(0, oo)n by
Yr = (n + 1 r)(xr Xr1) for 1 < r ::5 n,
YI = nxt,
with inverse Xr = L:k=l Yk/(n  k + 1) for r ~ 1. The Jacobian is (n !) 1 , whence the joint density
function of Yt, Y2, ... , Yn is
n.
i=l
k=l
34. Recall Problem (4.14.4). First, Zi = F(Xi), 1 ::::; i::::; n, is a sequence of independent variables
with the uniform distribution on [0, 1]. Secondly, a variable U has the latter distribution if and only
if  log U has the exponential distribution with parameter 1.
220
Solutions [4.14.35][4.14.37]
Problems
ltfollowsthatLi = log F(Xi), 1 ::S i ::S n, isasequenceofindependentvariables with the exponential distribution. The order statistics L(l) ... , L(n) are in order log F(X(n)), ... , log F(X(l)),
since the function log FO is nonincreasing. Applying the result of Problem (4.14.33), Et
n log F(X(n)) and
Er
1 < r ::S n,
log F(X(n+2r)) },
are independent with the exponential distribution. Therefore exp( Er ), 1 ::S r ::S n, are independent
with the uniform distribution.
35. One may be said to be in state j if the first j  1 prizes have been rejected and the jth prize has
just been viewed. There are two possible decisions at this stage: accept the jth prize if it is the best
so far (there is no point in accepting it if it is not), or reject it and continue. The mean return of the
first decision equals the probability j In that the j th prize is the best so far, and the mean return of the
second is the maximal probability f (j) that one may obtain the best prize having rejected the first j.
Thus the maximal mean return V (j) in state j satisfies
V(j)
= max{jln, f(j)}.
Now j In increases with j, and f(j) decreases with j (since a possible stategy is to reject the (j + l)th
prize also). Therefore there exists J such that j In ::=:: f (j) if and only if j ::S J. This confirms the
optimal stategy as having the following form: reject the first J prizes out of hand, and accept the
subsequent prize which is the best of those viewed so far. If there is no such prize, we pick the last
prize presented.
Let n, be the probability of achieving the best prize by following the above strategy. Let Ak be
the event that you pick the kth prize, and B the event that the prize picked is the best. Then,
Ln
rr, =
lP'(B
Ak)lP'(Ak) =
k=J+l
(k)(

Ln
k=J+l
1)
1
k 1 k
Ln
J
=
k=J+l
k ,
1
1
exp{ !(r 2  2A.x 2p,y 2vz
2
(2Jra ) 312
+ ;.2 + JL 2 + v2 )}
L:
Hm(x)Hn(x)(x) dx = (l)n
L:
L:
(l)n1
Hm(x)(n)(x) dx
H:n(x)(n1)(x)dx
221
[4.14.38][4.14.39]
Solutions
and the claim follows by the fact that H~m) (x) = m!.
(b) ~y Taylor's theorem and the first part,
tn
t)n
L Hn(x)
= L =cp(n)(x) = (x t),
oo
l/J(x)
oo (
n=O n!
n!
n=O
whence
38. The polynomials of Problem (4.14.37) are orthogonal, and there are unique expansions (subject
to mild conditions) of the form u(x) = 2::;~ 0 arHr(x) and v(x) = 2::;~ 0 brHr(x). Without loss of
generality, we may assume that E(U) = E(V) = 0, whence, by Problem (4.14.37a), ao = bo = 0.
By (4.14.37a) again,
00
00
L b?r!.
var(V) =
r=1
ifm=n,
ifm =/=n,
and so
= X(r)  X(r1) with the convention that X(Q) = 0 and X(n+1) = 1. By Problem
(4.14.21) and a change of variables, we may see that Y1, Y2, ... , Yn+1 have the distribution of a point
chosen uniformly at random in the simplex of nonnegative vectors y = (Y1, Y2, ... , Yn+ 1) with sum
1. [This may also be derived using a Poisson process representation and Theorem (6.12.7).] Consequently, the Yj are identically distributed, and their joint distribution is invariant under permutations
loo
[(I:
1
x 2n(1  x)n 1 dx
1= E
Yr) ] = (n
2
(n
+ 1)(n + 2)
+ 1)E(Yl) =
r=1
222
1, whence
Solutions [4.14.40][4.14.42]
Problems
implying that
I
JE(Y1 Y2) = (n + I)(n + 2),
and also
2
r(s + I)
JE(X(r)X(s)) = rlE(Y1 ) + r(s I)JE(Y1 Y2) = :(n+':I)::(n'+:2:)
40. (a) By paragraph (4.4.6), X 2 is f(!, !) and Y2 + z 2 is f(!, I). Now use theresultsofExercise
(4.7.I4).
(b) Since the distribution of x 2I R 2 is independent of the value of R 2 = x 2 + Y2 + z 2 ' it is valid
also if the three points are picked independently and uniformly within the sphere.
L:
2(x)<I>(A.x) dx =
=
L:
L:
L:
(x)dx +
because the second integrand is odd. Using the symmetry of, the density function of IYI is
(x) + (x){ <I>(A.x) <1>(J...x)} + (x) + (x){ <1>(A.x) <I>(A.x)} = 2(x).
(c) Finally, make the change of variables W = IYI. Z =(X+ A.IYI)/)I + A. 2, with inverse IYI = w,
x = zJI + J...2 A.w, and Jacobian )I+ A. 2. Then
Jw,z(w,
+ A. 2  A.w, w)
> 0,
R.
The result follows by integrating over w and using the fact that
i), V1, V2 are N(O, ~),and U1, U2, V1, V2 are independent. The answer is
I)
where the
Ni
= lP'(B ::::; 4) = 4
223
x2(2)
[4.14.43][4.14.48]
Solutions
where we have used the result of Exercise (4.7.14), and B is a betadistributed random variable with
parameters 1, 1.
43. The argument of Problem (4.14.42) leads to the expression
Vf +
vf + vf)
= lP'(K 1 :::
1K2)
.J3
3 4n'
1
1
= JP'(B ::; 4 ) =
(b) var(Sn) = na 2 and lE[(Sn  np,) 3 ] = nlE[(X 1  p,) 3 ] plus terms which equal zero because
lE(X1 p,) = 0.
(c) If Y is Bernoulli with parameter p, then skw(Y) = (1  2p)/ .fii(j, and the claim follows by (b).
(d) m1 =A, m2 =A+ A2, m3 = A3 + 3A 2 +A, and the claim follows by (a).
(e)SinceAXisf(l, t), wemayaswellassumethatA = 1. ItisimmediatethatlE(Xn) = f(t+n)/f(t),
whence
t(t + l)(t + 2) 3t. t(t + 1) + 2t 3
2
skw(X) =
t 312
= 0 .
45. We find as above that kur(X) = (m4 4m3m1 + 6m2mi 3mj)/a 4 where mk = lE(Xk).
+
oo that
x)n
+ eex.
1
1
lE(X(2)) = +  1 ,
n
nwhence, by Lemma (4.3.4),
!oo
oo
n+oo
1 + 1 + + 1logn ) = y.
n
n 1
47. By the argument presented in Section 4.11, conditional on acceptance, X has density function
fs. You might use this method when fs is itself particularly tiresome or expensive to calculate. If
a(x) and b(x) are easy to calculate and are close to fs, much computational effort may be saved.
48. M = max{U1, U2, ... , Uy} satisfies
e1  1
JP'(M::; t) = lE(ty) =   .
e I
224
Solutions [4.14.49][4.14.51]
Problems
Thus,
lP'(Z ~
z) = lP'(X
=
Y ~
(e l)eLzJ2
1 e1
+ (e
LzJ + 1 z)
LJ 1 elzJ+1z 1
.
e 1
= ez
l)e z 
49. (a) Y has density function eY for y > 0, and X has density function fx(x) = aeax for x > 0.
Now Y ~ (X  a) 2 if and only if
I 2
which is to say that aVfx(X) ~ f(X), where a = a 1e2" 01 .J2[ii. Recalling the argument of
Example (4.11.5), we conclude that, conditional on this event occurring, X has density function f.
(b) The number of rejections is geometrically distributed with mean a 1, so the optimal value of a is
I 2
= 1.
1}
!
with probability
with probability
conditional on Y >
1(X 
a )2 ,
fo
~du = n/4.
In
(b) lE(Y) = 
n lo
51. You are asked to calculate the mean distance of a randomly chosen pebble from the nearest
collection point. Running through the cases, where we suppose the circle has radius a and we write
P for the position of the pebble,
(i)
lEIOPI =
(ii)
lEIAPI
__!__2
na
lE(IAPI
(iii)
IBPI)
=
fa r 2dr dB
= 2a.
lo!n lo2acos8
na
{ 2n
Jo Jo
r 2 drdB
32a
= .
9n
lo lo
4a { 16
3rr 3
l!n lo
~
1
~ }
 617 v2+
2 1og(l +v2)
:::::::
2a
x 1.13.
(iv)
lE(IAPI
r
r 2dr dB+ [!n f 2acose r 2dr dB}
lo
J!n lo
[!n
lo
1
= 2a {lojn 3.J3
cosec 3 (n
 +e)
7r
= 2a { 16 7r
dB+
.!...!.,J3
+ 3.J3 log~} ::::::: 2a
4
16
2
3
225
~!n 8 cos 3 B dB }
!n
x 0.67.
[4.14.52][4.14.55]
Solutions
52. By Problem (4.14.4 ), the displacement of R relative to P is the sum of two independent Cauchy
random variables. By Exercise (4.8.2), this sum has also a Cauchy distribution, and inverting the
transformation shows that e is uniformly distributed.
~ occurs if and only if
the sum of any two parts exceeds the length of the third part.
(a) If the breaks are at X, Y, where 0 < X < Y < 1, then ~occurs if and only if 2Y > 1, and
2(Y  X) < 1 and 2X < 1. These inequalities are satisfied with probability
53. We may assume without loss of generality that R has length 1. Note that
i.
(b) The length X of the shorter piece has density function fx (x) = 2 for 0 ~ x ~ ~. The other pieces
are oflength (1 X)Y and (1 X)(l Y), where Y is uniform on (0, 1). The event~ occurs if and
only if2Y < 2XY + 1 and X+ Y XY >
and this has probability
i.
lo0 ~ { 2(1 1 X) 
1 2x } dx = log(4/e).
2(1  X)
io X), iO X), where X is uniform on (0, 1). The event~ occurs
lx
2
2 (1 X)
.fi 
>
;;:;
v2
1. Hence,
IP'(obtuse
IM =
r) = IP' ( X1X
i Hence,
r ) = 2r
,
l+r
226
Solutions [4.14.56][4.14.59]
Problems
I
function F(z) = 2../i z and density function fz(z) = z2"  1, for 0:::::: z:::::: 1. Therefore R 2 has
the density f given by
f(r) =
1
r (1 _ 1) (~
_ 1) dz
{ lo 1 ../i
1 (../i
1  1) ( 1
 1) dz
~
r1
if0::S:r::S:1,
if 1 ::S: r ::S: 2.
b11  dz = 2 (.
sm
a ../i~
1 ~ .1l)
  sm
r
56. We use an argument similar to that used for Buffon's needle. Dropping the paper at random
amounts to dropping the lattice at random on the paper. The mean number of points of the lattice in a
small element of area dAis dA. By the additivity of expectations, the mean number of points on the
paper is A. There must therefore exist a position for the paper in which it covers at least rA l points.
57. Consider a small element of surface dS. Positioning the rock at random amounts to shining light
at this element from a randomly chosen direction. On averaging over all possible directions, we see
that the mean area of the shadow cast by d S is proportional to the area of d S. We now integrate over
the surface of the rock, and use the additivity of expectation, to deduce that the area A of the random
shadow satisfies E(A) = cS for some constant c which is independent of the shape of the rock. By
considering the special case of the sphere, we find c =
It follows that at least one orientation of
the rock gives a shadow of area at least S.
;! .
;!
58. (a) We have from Problem (4.14.1lb) that Yr = Xr/(X1 ++X,) is independent of X1 +
+X,, and therefore of the variables Xr+1 Xr+2 , Xk+1 X1 + + Xk+1 Therefore Yr is
independent of {Yr+s : s ~ 1}, and the claim follows.
(b) Lets= x1 + ... + xk+1 The inverse transformation X1 = Z1S, xz = zzs, ... ' Xk = ZkS,
xk+1 = s  z1s zzs  ZkS has Jacobian
s
0
0
Z1
zz
=sk.
1=
Zk
s
s
s
1 Z1  Zk
{IT
r=1
f'(fJr)
f'(fJk+J)
k
f(A.,
f3, s)
(IT
zf' 1) (1  Z1   Zk)fJk+J 1,
r=1
where
59. Let C = (crs) be an orthogonal n x n matrix with Cni = 1/ ,fii for 1 ::S: i ::S: n. Let Yir =
2::~= 1 XisCrs. and note that the vectors Y, = (Y1, Yz, ... , Ynr), 1 ::S: r ::S: n, are multivariate
normal. Clearly EYir = 0, and
E(YirYjs) = LCrtCsuE(XitXju) = LCrtCsu8tuVij = LCrtCstVij = 8rsVij
t,u
t,u
t
227
[4.14.60][4.14.60]
Solutions
where Dtu is the Kronecker delta, since C is orthogonal. It follows that the set of vectors Y r has the
same joint distribution as the set of Xr. Since C is orthogonal, Xir = 2::~= 1 Csr Yis, and therefore
n1
= LYisYjs YinYjn = LYisYjs
s=1
This has the same distribution as Tij because the Yr and the X7 are identically distributed.
60. We sketch this. Let EIPQRI = m(a), and use Crofton's method. A point randomly dropped in
S(a + da) lies in S(a) with probability
2
2da
( a ) = 1  +o(da).
a+da
a
Hence
dm
6m
6mb
=+,
da
a
a
where mb(a) is the conditional mean of IPQRI given that Pis constrained to lie on the boundary of
S(a). Let b(x) be the conditional mean of IPQRI given that Plies a distance x down one vertical
edge.
T1
R1
T2
R2
By conditioning on whether Q and R lie above or beneath P we find, in an obvious notation, that
x)2
(ax) 2
b(x) = ( ~ mR 1 + amR2
+ 2x(ax)
a2
mR 1,Rz
By Exercise (4.13.6) (see also Exercise (4.13.7)), mR 1 ,R2 = icia)(ia) = ~a 2 . In order to find
m R 1, we condition on whether Q and R lie in the triangles T1 or T2, and use an obvious notation.
i,
m R!
141
4 'I7 2 ax
1
143
13
+ 41 4
'I7 zax + 2 9 gax = TOSax.
x _ (::_)213ax
b( )  a
108
228
12x 2
108
Solutions [4.14.61][4.14.63]
Problems
11oa b(x) dx =
mb(a) = 
11a 2
.
108
We substitute this into the differential equation to obtain the solution m(a) = {;ha 2 .
Turning to the last part, by making an affine transformation, we may without loss of generality
take the parallelogram to be a square. The points form a convex quadrilateral when no point lies
inside the triangle formed by the other three, and the required probability is therefore 1  4m (a) I a 2 =
1 ~ = ~
61. Choose four points P, Q, R, S uniformly at random inside C, and let T be the event that their
convex hull is a triangle. By considering which of the four points lies in the interior of the convex
hull of the other three, we see that IP'(T) = 41P'(S E PQR) = 4EIPQRIIICI. Having chosen P, Q, R,
the four points form a triangle if and only if S lies in either the triangle PQR or the shaded region A.
Thus, IP'(T) ={I AI+ EIPQRI}IICI, and the claim follows on solving for IP'(T).
62. Since X has zero means and covariance matrix I, we have that E(Z) = fL + E(X)L = fL, and the
covariance matrix of Z is E(L'X'XL) = L'IL = V.
63. Let D = (dij) = ABC. The claim is trivial if D = 0, and so we assume the converse. Choose
i, k such that dik i= 0, and write Yi = L,}=t dijXj = S + dikXk Now IP'(Yi = 0) = E(IP'(xk =
 SI dik I S)). For any given S, there is probability at least ~ that Xk i=  S I dik. and the second claim
follows.
Let Xt, x2, ... , Xm be independent random vectors with the given distribution. If D i= 0, the
probability that Dxs = 0 for 1 :::; s :::; m is at most ( ~ )m, which may be made as small as required by
choosing m sufficiently large.
229
(a)lflsl<(1p) 1,
G(s)
~ sm
= !;:o
(n + mm  1) pn(l 
p)m
1 s(1 p)
}n .
Therefore the mean is G'(l) = n(1  p)/ p. The variance is G"(l) +G' (1) G'(1) 2 = n(l  p)f p 2 .
(b) If lsi < 1,
G(s)
sm
m= 1
Therefore G' (1) = oo, and there exist no moments of order 1 or greater.
(c) If p <lsi< p 1,
G(s) =
m~oo
00
sm
1 p ) plml = 1 p
1+ P
1+ p
1 + _!_!!__ +
pfs
1 sp
1 (pjs)
T (1)
{(1s)G'(s)+1G(s)}
= hm
2
st1
(1s)
Solutions [5.1.3][5.1.5]
Generating functions
(i) We have that Gx,y(s, t) = E(sXtY), whence Gx,y(s, 1) = Gx(s) and Gx,Y(l, t) = Gy(t).
a2
E(XY) = E (xYsxItYI) I
= Gx,y(s,
t) I
.
s=t=l
as at
s=t=l
4.
(a)
G(s, t) =
jl
j=Ok=O
if f31tl < 1
1
f3t }
(1f3t)f3
1(as/{3)1ast
(1  a)(f3 a)
(1  ast)(f3 as)
(1  a)(f3 a) {
if~lsl
f3
< 1
(the condition alstl < 1 is implied by the other two conditions on s and t). The marginal generating
functions are
G(s, 1) = (1 a)(f3 a) , G(l, t) = 1a
,
1 at
(1  as)(f3 as)
and the covariance is easily calculated by the conclusion of Exercise (5.1.3) as a(l  a) 2.
(b) Arguing similarly, we obtain G(s, t) = (e 1)/{e(l tes 2)} if itles 2 < 1, with marginals
G(s,l)=
1 e 1
,
1 es 2
G(l,t)=
1 e 1
,
1 te 1
if ltp(l  p
+ sp)l
< 1.
5.
log{ 1  p + p 2 (1  s)}
log(l _ p)
,
p 2 {p + log(l  p)}
(1 p)2{log(l p)}2.
231
[5.1.6][5.2.1]
Solutions
where p + q = 1.
(ii) More generally, if each toss results in one oft possible outcomes, the ith of which has probability
Pi, then the corresponding quantity is a function of t variables, x1, x2, ... , x 1, and is found to be
(p1x1 + P2X2 + + PtXt)n.
6.
We have that
JE(sx)
= lE{lE(sx I U)} =
loo
{1 + u(s 1)}n du
=  n+1
sn+1
1s
the probability generating function of the uniform distribution. See also Exercise (4.6.5).
7.
We have that
Gx,Y,z(x, y,z)
= G(x, y,z, 1) =
=
= Gx(x)Gy(y)Gz(z),
whence X, Y, Z are independent. The same conclusion holds for any other set of exactly three random
variables. However, G(x, y, z, w) i= Gx(x)Gy(y)Gz(z)Gw(w).
8. (a) We have by differentiating that JE(X 2n) = 0, whence JP(X = 0) = 1. This is not a moment
generating function.
(b) This is a moment generating function if and only if Er Pr = 1, in which case it is that of a random
variable X with JP(X = ar) = Pr
9. The coefficients of sn in both combinations of G 1, G2 are nonnegative and sum to 1. They are
therefore probability generating functions, as is G (as) f G (a) for the same reasons.
T(s)
~
L smJP(X 2: m)
= 1 +s L
m=O
= 1+
s(1 G(s))
1 s
k=O
Now,
so that
T(s) T(O)
s
= G s (0) =
1. Therefore
i=1
j=1
~
lP(X 2: i) =LSi
.
]=I
(j 1) ..
.
1
(1)1 1 ,
232
1 _::: i _::: n.
1 sG(s)
1 s
Solutions [5.2.2][5.2.4]
Some applications
Similarly,
Gs(s) Gs(O)
s
whence the second formula follows.
+ s) T(O)
T(l
1 +s
2. Let Ai be the event that the ith person is chosen by nobody, and let X be the number of events
A 1, A2, ... , An which occur. Clearly
j)
nn 1
zJ
12
lJ
j (
n j n 1
1)
n j
if i 1 ::j= i2 ::j= ::j= ij, since this event requires each of i 1, ... , ij to choose from a set of n  j people,
and each of the others to choose from a set of size n j  1. Using Waring's Theorem (Problem
(1.8.13) or equation (5.2.14)),
IP'(X = k) = t<1)jk
. k
]=
where
Sj =
(~)
1
(j)sj
k
(nj)j (nj1)nj.
n1
n1
n
( 1 = L(1)jk
. k
]=
while IF'( X
3.
0)
1)
k 1
Sj,
1_:::k_:::n,
= 1.
(a)
E(xX+Y)
G(s) =
{s(l p)}k
k= 1 klog(ljp)
Gy(s)
4.
= GN(G(s)) = et.L(G(s) 1) =
)JL/1ogp
1s(1p)
Clearly,
E ( 1 ) =E
1+ X
1 (1  'A/n)n+ 1
1 )
1 +X =
'A(n + 1)/n
233
+ o(l)
+
1 eJ...
'A
1) }.
[5.2.5][5.3.1]
Solutions
5.
H (s)
1 qs
(1s){l(qp)s}
1
1
}
= 21 {  + 1(qp)s
1s
6. By considering the event that HTH does not appear inn tosses and then appears in the next three,
we find that IP'(X > n) p 2 q = IF'( X = n + 1) pq +IF'( X = n + 3). We multiply by sn+ 3 and sum over
n to obtain
1 E(sx)
:''p 2 qs 3 = pqs 2 E(sx) +E(sx),
1s
which may be solved as required. Let Z be the time at which THT first appears, soY= min{X, Z}.
By a similar argument,
IP'(Y > n)p2q = IP'(X = Y = n + 1)pq + IP'(X = Y = n + 3) + IP'(Z = Y = n + 2)p,
IP'(Y > n)q 2 p = IP'(Z = Y = n + 1) pq + IP'(Z = Y = n + 3) +IF'( X = Y = n + 2)q.
We multipy by sn+ 1, sum over n, and use the fact that IP'(Y = n) =IF'( X= Y = n) + IP'(Z = Y = n).
7.
Suppose there are n + 1 matching letter/envelope pairs, numbered accordingly. Imagine the
envelopes lined up in order, and the letters dropped at random onto these envelopes. Assume that
exactly j + 1 letters land on their correct envelopes. The removal of any one of these j + 1 letters,
together with the corresponding envelope, results after reordering in a sequence of length n in which
exactly j letters are correctly placed. It is not difficult to see that, for each resulting sequence of length
n, there are exactly j + 1 originating sequences oflength n + 1. The first result follows. We multiply
by sj and sum over j to obtain the second. It is evident that G 1(s) = s. Either use induction, or
integrate repeatedly, to find that Gn(s) = L:~=0 (s W /r!.
8.
9.
I A)}= E(eA(s1)) =
_IL_ f (s)k
fL
=
fL (s  1)
fL + 1 k=O
fL + 1
Since the waiting times for new objects are geometric and independent,
iCi)k4 }, fork~ 4.
Let Ak be the event that the walk ever reaches the point k. Then Ak 2 Ak+ 1 if k
r1
IP'(M ~ r) = IP'(Ar) = IP'(Ao)
IJ IP'(Ak+l I Ak) =
k=O
234
(p/q/,
0,
0, so that
Solutions
Random walk
since lP'(Ak+1
2.
I Ak)
[5.3.2][5.3.4]
00
""'"'
L.J s 2k 2kfo(2k)
00
S
= s F0I (s) = ;;:z
= s 2 lP'o(s) = ""'"'
L.J s 2k Po(2k 2),
k=1
1  s~
k=1
an=
fo(k),
k=2n+2
keven
with the convention that a 0 = 1. We have used the fact that ultimate return to 0 occurs with probability
1. This sequence has generating function given by
00
00
L s2n L
n=O
!k1
00
fo(k) =
k=2n+2
k even
fo(k)
k=2
k even
1  Fo(s)
s2n
n=O
1 s 2
by Theorem (5.3.1c)
00
= lP'o(s) = L
s2nlP'(S2n
= 0).
n=O
Now equate the coefficients of s 2n. (Alternatively, use Exercise (5.1.2) to obtain the generating
function of the an directly.)
3. Draw a diagram of the square with the letters ABCD in clockwise order. Clearly p AA (m) = 0 if
m is odd. The walk is at A after 2n steps if and only if the numbers of leftward and rightward steps are
equal and the numbers of upward and downward steps are equal. The number of ways of choosing
2k horizontal steps out of 2n is (~k). Hence
Writing FA (s) for the probability generating function of the time T of first return, we use the
argument which leads to Theorem (5.3.1a) to find that GA(s) = 1 + FA(s)GA(s), and therefore
FA(s) = 1GA(s) 1.
4. Write (Xn, Yn) for the position of the particle at time n. It is an elementary calculation to show
that the relations Un = Xn + Yn, Vn = Xn Yn define independent simple symmetric random walks
U and V. Now T = min{n: Un = m}, and therefore Gr(s) = {s 1(1 ~)}m for lsi :::: 1
by Theorem (5.3.5).
235
[5.3.5][5.3.6]
Solutions
where we have used the independence of U and V. This converges if and only if I~ (s + s I) I _:s 1,
which is to say that s = 1. Note that GT (s) converges in a nontrivial region of the complex plane.
5.
Let T be the time of the first return of the walk S to its starting point 0. During the timeinterval
(0, T), the walk is equally likely to be to the left or to the right of 0, and therefore
Lzn = {
+ L'
TR
2nR
if T _:s 2n,
ifT>2n,
where R is Bernoulli with parameter~, L' has the distribution of LznT, and Rand L' are independent.
It follows that Gzn (s) = JE(s Lzn) satisfies
n
where f(2k) = IP'(T = 2k). (Remember that Lzn and T are even numbers.) Let H(s, t)
L:~o t 2n Gzn (s). Multiply through the above equation by t 2n and sum over n, to find that
H(s, t) = ~H(s, t) {F(t) + F(st)} + ~ {J(t) + J(st)}
J(x)=L::x 2nLf(2k)=
n=O
k>n
~
1
lxl
< 1,
by the calculation in the solution to Exercise (5.3.2). Using the fact that F(x) = 1  ~.we
deduce that H(s, t) = 1/VO t 2)(1 s2t2). The coefficient of s 2kt 2n is
IP'(Lzn=2k)= (n_!k)(l)nk.
(~~)(1)k
2)Zn =
2k) (2n= ( k
n _ k2k) ( 1
6. We show that all three terms have the same generating function, using various results established
for simple symmetric random walk. First, in the usual notation,
oo
2m
'
2s2
L 4mlP'(Szm = O)s
= 2sP0 (s) =
2 312 .
m=O
(1 s )
Secondly, by Exercise (5.3.2a, b),
m
lE(T
1\
236
Solutions [5.3.7][5.3.8]
Random walk
Hence,
oo 2m
s2Po(s)
'
L
s E(T A 2m)=  2 +sP0 (s) =
m=O
1 s
2s2
2 312 .
(1  s )
Finally, using the hitting time theorem (3.10.14), (5.3.7), and some algebra at the last stage,
oo
m=O
oo
s 2m2EIS2m I = 4
oo
s 2m
m=O
L 2k1P'(S2m = 2k) = 4 L
s 2m
m=O
k=1
L 2mf2k(2m)
k=1
d ~ 2m~
d
F1(s) 2
2s 2
= 4s L s L hk(2m) = 4s2 =
2 312 .
ds m=O
k= 1
ds 1  F 1(s)
(1  s )
7. Let In be the indicator of the event {Sn = 0}, so that Sn+1 = Sn + Xn+1 +ln. In equilibrium,
E(So) = E(So) + E(X 1) + E(/o), which implies that IP'(So = 0) = E(/o) = E(X 1) and entails
E(X t) ::::: 0. Furthermore, it is impossible that IP'(So = 0) = 0 since this entails IP'(So =a) = 0 for all
a < oo. Hence E(X 1) < 0 if Sis in equilibrium. Next, in equilibrium,
Now,
E{ zSn+Xn+l +In (1  In)} = E(zSn
I Sn
Hence
E(zsO) = E(zxi) [{E(zso)  IP'(So = 0)} +ziP'( So = 0)]
which yields the appropriate choice for E(zsO).
8.
The hitting time theorem (3.10.14), (5.3.7), states that IP'(Tob = n) = (lbl/n)IP'(Sn =b), whence
E(Tob
I Tob
< oo)
L IP'(Sn =b).
I Tob
The required conditional mean may be found by conditioning on the first step, or alternatively
as follows. Assume first that p < q, so that IP'(Tob < oo) = (p/q)b by Corollary (5.3.6). Then
L:n IP'(Sn =b) is the mean of the number N of visits of the walk to b. Now
IP'(N=r)=(~rpr 1 (1p),
r0::1,
where p = IP'(Sn = 0 for some n 0:: 1) = 1  IP ql. Therefore E(N) = (pfq)b liP ql and
E(Tob I Tob < oo)
b
(p/q)b
(p/q)b IP _ ql.
We have when p > q that IP'(Tob < oo) = 1, and E(TOI) = (p q) 1. The result follows from
the fact that E(Tob) = bE(To1).
237
[5.4.1][5.4.4]
Solutions
= fLnmE(Z~) E(Zm)E(Zn) =
f.lnm var(Zm),
2. Suppose 0 :::0 r :::0 n, and that everything is known about the process up to timer. Conditional on
this information, and using a symmetry argument, a randomly chosen individual in the nth generation
has probability 1/ Zr of having as rth generation ancestor any given member of the rth generation.
The chance that two individuals from the nth generation, chosen randomly and independently of each
other, have the same rth generation ancestor is therefore 1/Zr. Therefore
lP'(L < r)
= E{lP'(L
< r
I Zr)} = E(1 z; 1)
and so
lP'(L
= r) = lP'(L
< r
+ 1) lP'(L
< r)
0:::0 r < n.
If 0 < lP'(ZI = 0) < 1, then almost the same argument proves that lP'(L
Y/r Ylr+l forO :::0 r < n, where Y/r = E(Z; 1 I Zn > 0).
3.
Zn > 0)
lP'(Zn
= 0) = Gn(O) =
{n:
if p = q,
q (p n q n)
pn+l _ qn+l
if p # q,
= n) = lP'(Zn = 0) lP'(Zn1 = 0) =
{ n(n
+ 1)
p
n1 n
2
q (p q)
if p
= q,
if p # q.
4.
(a) As usual,
This suggests that Gn(s) = 1  al+.B++.BnI (1  s).Bn for n :::: 1; this formula may be proved
easily by induction, using the fact that Gn(s) = G(Gn1 (s)).
(b) As in the above part (a),
238
Solutions
[5.4.5][5.5.1]
5.
Let Zn be the number of members of the nth generation. The (n + 1)th generation has size
Cn+l + ln+l where Cn+I is the number of natural offspring of the previous generation, and ln+l is
whence
Gn+I(s)
6.
By Example (5.4.3),
lE(s n) =
Differentiate and set s
n  (n  1)s
n 1
1
=  +
,
n + 1 ns
n
n2(1 + n1  s)
n:::: 0.
= 0 to find that
Similarly,
oo
oo
00
n2
00
oo
1 2
oo
lE(V2)
lE(V3)
(n + 1) 2  2(n + 1) + 1
(n + 1)4
1 2
1 4
()JT + 90JT  2
00
G(Gtu(s))fr(u) du +
00
sfr(u) du.
a
Gt(s)
= G(Go(s))fr(t) +
at
 {G(Gtu(s))}
at
at
a
=a
{G(Gru(s))},
u
239
2:: (n+1).,3.
n=O
[5.5.2][5.5.2]
Solutions
1
00
=A {G(Gt(s)) G 1 (s)}.
ift = A(G 12 
Gt)
with boundary condition Go(s) = s. Integrate to obtain At+ c(s) = log{l  G( 1} for some function
c(s). Using the boundary condition att = 0, we find thatc(s) = log{l G01} = log{1s 1}, and
hence G 1 (s) =seAt /{1 s(1 eAt)}. Expand in powers of s to find that Z(t) has the geometric
distribution JID(Z(t) = k) = (1 eAt)k 1eA.t fork 2: 1.
2.
ift
with boundary condition Go(s)
1
= 2(1
+ G21 )  Gt
4/t
G t (s ) _ 2s + t(l  s) _ ,,,2+t(1s)
2+t(ls)
2t
)n ,
n 2: 1,
and therefore
( () )?:;
JP>Zt >k
00
4  ( tt(2 + t) 2 + t
)n  2t ( 2 +tt )k ,
JID(Z(t) 2: xt
I Z(t)
> 0)
JID(Z(t) 2: 1)
t ) LxtJ1
2+t
= 
240
2)
= 1 +t
1LxtJ
+ e 2x.
Solutions
Characteristic functions
[5.6.1][5.7.2]
3.
4.
= L:~ 1 IXi I are such that IZn I :::: Z, and the result follows by
n>oo
= limsuplE(Xn).
n>oo
ur dF(u)+ 0
as X+
00,
Jrx,oo)
The first term on the righthand side is negative. The integrand in the second term satisfies sus  11P'(I X I >
u) :::: susl ur for all large u. Therefore the integral is bounded uniformly in M, as required.
5. Suppose first that, for all E > 0, there exists 8 = 8(E) > 0, such that lE(IXI/A) < E for all A
satisfying IP'(A) < 8. FixE > 0, and find x (> 0) such that IP'(IXI > x) < 8(E). Then, for y > x,
Hence
2.
(i) It is the case that Re{(t)} = lE(cos t X), so that, in the obvious notation,
Re{1  (2t)}
L:
L:
:::: 4
=2
L:
= 4Re{1
241
[5.7.3][5.7.6]
Solutions
(ii) Note first that, if X and Y are independent with common characteristic function , then X  Y
has characteristic function
3.
r=l
ExpandS(()/ in powers of e, and equate the coefficients of e, e 2 , e 3 , in tum, to find thatk1 (X) = mt.
k2(X) = m2 mr, k3(X) = m3 3mtm2 + 2mj.
(b) If X and Yare independent, Kx+y(()) = log{lE(e8 X)lE(e 8 f)} = Kx(())
claim is immediate.
I 2
4.
5.
since only numbers of the form x = a + bm make nonzero contributions to the sum.
Suppose in addition that X has span b, and that lc/>x(T)I = 1 for some T E (0, 2nlb). Then
c/>x(T) = eic for some c E JR. Now
JE(cos(TX _c))= ~lE(eiTXic
+ eiTX+ic)
= 1,
using the fact that 1E(eiTX) = cJ>x(T) = eic. However cosx :;:; 1 for all x, with equality if and
only if x is a multiple of 2rr. It follows that T X cis a multiple of 2rr, with probability 1, and hence
that X takes values in the set L(ciT, 2rr IT). However 2rr IT > b, which contradicts the maximality
of the span b. We deduce that no such T exists.
(b) This follows by the argument above.
6.
This is a form of the 'RiemannLebesgue lemma'. It is a standard result of analysis that, forE > 0,
there exists a step function gE such that f~oo If (x)  gE (x) I dx < E. Let E (t) = f~oo eitx gE (x) dx.
Then
If we can prove that, for each E, lc/>E(t)l + 0 as t+ oo, then it will follow that lc/>x(t)l < 2E for all
large t, and the claim then follows.
242
Solutions [5.7.7][5.7.9]
Characteristic functions
Now gE(x) is a finite linear combination of functions of the form ciA (x) for reals c and intervals
A, that is gE(x) = 'LJ=l q!Ak (x); elementary integration yields
2 K
II/IE (t) I < 
2:: Ck + 0,
as t + oo.
 t k=l
7.
M 2(s)
X
= lE(esX 2 ) =
100
_ 00
1
I (x JL )2 dx
esx 2 e2
..(iii
.J[=2S
112s ) ,
exp ( 1  2s
My(s) =
ll
n {
.Jf=2S
11~s
1
exp (  1 ) } =
exp ( se ) .
1  2s
(1  2s)nf2
1  2s
It is tempting to substitute s = it to obtain the answer. This procedure may be justified in this case
using the theory of analytic continuation.
x2 (n).
Hence T 2 is
(A 2 + B)jm
Z= ''Vjn
where A, B, V are independent, A being N(.fO, 1), B being x2 (m1), and V being x2 (n). Therefore
lE(Z) =
=
where we have used the fact (see Exercise (4.10.2)) that the F(r, s) distribution has mean sj(s 2)
if s > 2.
9. ~t X be independent of X with the same distribution. Then 11/11 2 is the characteristic function of
X  X and, by the inversion theorem,
1
2n oo
oo
X be differentiable at 0.
[5.7.10][5.8.2]
Solutions
10. By definition,
eityt/Jx(y)
j_:
eiy(xt) fx(x)dx.
Now multiply by fy (y ), integrate over y E JR, and change the order of integration with an appeal to
Fubini's theorem.
J:
11. (a) We adopt the usual convention that integrals of the form
g(y) dF(y) include any atom of
the distribution function Fat the upper endpoint v but not at the lower endpoint u. It is a consequence
that Fr: is rightcontinuous, and it is immediate that Fr: increases from 0 to 1. Therefore Fr: is a
distribution function. The corresponding moment generating function is
Mr:(t)
00
etx dFr:(x)
oo
= 1M(t)
00
etx+r:x dF(x)
oo
M(t + r) .
M(t)
+ r)
Mx(t
Mx+y(t)
+ r)My(t + r)
Mx(t)My(t)
the product of the moment generating functions of the individual tilted distributions.
l(t)l 2 .
(iv) Let X have characteristic function , and let Z be equal to X with probability
otherwise. The characteristic function of Z is given by
1and to X
1,
(v) If X is Bernoulli with parameter then its characteristic function is (t) = ~ + eit. Suppose Y
is a random variable with characteristic function 1/f(t) = l(t)l. Then 1/f(t) 2 = (t)(t). Written
in terms of random variables this asserts that Y1 + Y2 has the same distribution as X 1  Xz, where
the Y; are independent with characteristic function 1/f, and the X; are independent with characteristic
function. Now Xj E {0, 1}, so that Xr Xz E {1, 0, 1}, and therefore Yj E
Write
= lP'(Yj =
{1, 1l
J). Then
+ Yz = l) = a 2 = lP'(Xr Xz = 1) = ~.
lP'(Yr + Yz = 1) = (1 a) 2 = lP'(Xr Xz = 1) =
lP'(Yr
Fort~
0,
~.
0.
244
Solutions [5.8.3][5.8.5]
Mz(t)
oo (v _ l)mn1
Imn =
(A.v  t)m
dv
satisfies
1
(vl)mn 1 ] 00
Imn =[ A.(m 1) (A.v t)m1 1
mn1
A.(m 1) Im1,n =cmnA.I
( , , ) m1,n
c'
for somec' depending onm, n, A.. Therefore Mz(t) = c" (A. t)n for some c" depending on m, n, A..
However Mz(O) = 1, and hence c" = A.n, giving that Z is r(A., n). Throughout these calculations
we have assumed that tis sufficiently small and positive. Alternatively, we could have set t =is and
used characteristic functions. See also Problem (4.14.12).
4.
We have that
1E(e 11. x2 ) =
00
oo
. 2
1  exp (  (x  ~t) 2 ) dx
ettx
2
J2na2
2a
= oo
1
(
exp oo J2na2
y'1 2a2it
exp (
t) (
itf/,2 )
.
dx
1  2a 2zt
exp
itf/,2 ) .
1 2a 2it
The integral is evaluated by using Cauchy's theorem when integrating around a sector in the complex
plane. It is highly suggestive to observe that the integrand differs only by a multiplicative constant
from a hypothetical normal density function with (complex) mean ~t(1  2a 2 it) 1 and (complex)
variance a 2 (1 2a 2 it) 1.
5.
= 0 and a 2
x2 ( 1) distribution.
245
[5.8.6][5.8.6]
Solutions
Now
E(exp{it2/X~}) =1oo
00
There are various ways of evaluating this integral. Using the result of Problem (5.12.18c), we find
that the answer is eltl, whence X 1/ Xz has the Cauchy distribution.
(d) We have that
E(eitX1X2) = E{E(eitX1X2 I Xz)} = E(if>x 1(tXz)) =
=
E(e~t 2 X~)
~exp{ix 2 (1 +t 2 )}dx = ~,
1 + t2
00
oo v2rr
on observing that the integrand differs from the N (0, (1 + t 2 ) 2) density function only by a multiplicative constant. Now, examination of a standard work of reference, such as Abramowitz and Stegun
(1965, Section 9.6.21), reveals that
!o
oo cos(xt) d
~
y 1 r r
( )
t=Kox,
where Ko(x) is the second kind of modified Bessel function. Hence the required density, by the
inversion theorem, is f(x) = Ko(lxl)jn. Note that, for small x, Ko(x) ~ logx, and for large
positive x, Ko(x) ~ex .Jnx/2.
As a matter of interest, note that we may also invert the more general characteristic function
if>(t) = (1 it)a(l + it)f3. Setting 1 it= zjx in the integral gives
1 1oo
f(x) = 2rr
eitx
exxaI 1x+ixoo
ez dz
dt = ;;;oo (1 it)a(l + it)f3
2f32ni
xixoo (z)a (1 + z/(2x))f3
1
ex (2x) z;(f3a)
''wz;(af3),
1
1
(2x)
f(a)
2 (Iaf3)
where W is a confluent hypergeometric function. When a = f3 this becomes
I
f(x)
(xj2)az;
f(a).J]r Ka~ (x)
iW
6. The joint characteristic function of X= (X 1, Xz, ... , Xn) satisfies if>x(t) = E(eitX') = E(eiY)
where t = (tJ, tz, ... , tn) E JRn andY= tX' = t1X 1 + + tnXn. Now Y is normal with mean and
variance
n
var(Y) =
j,k=I
j=l
where fL is the mean vector of X, and Vis the covariance matrix of X. Therefore if>x(t) = if>y(l) =
exp(itp/ itVt') by (5.8.5).
246
Solutions [5.8.7][5.9.2]
Let Z = X JL. It is easy to check that the vector Z has joint characteristic function Jz(t) =
I 2
By definition, E(eitX)
loo
oo
= E(cos(tX)) + iE(sin(tX)).
A2
cos(tx )Ae Ax dx =  2,
2 A +t
loo
By integrating by parts,
oo .
sm(tx)Ae
Ax
dx
At
= 2 2,
A +t
and
A2 +iA.t
A
A2 + t 2 = A  it .
9.
(a) We have that elxl =ex l{x:c:O) +ex l{x<O) whence the required characteristic function is
J(t)
1(
1)
(b) By a similar argument applied to the f(l, 2) distribution, we have in this case that
1(
J(t) =
1
(1  it)2
+ (1 + it)2
1 t 2
= (1
+ t2) 2 .
10. Suppose X has moment generating function M(t). The proposed equation gives
M(t)
lo
M(ut) 2 du
+ t).
= E(eisX+itY) = JsX+tY0).
The fact that c/Jx, y may be expressed in terms of the characteristic function of a single normal variable
is sometimes referred to as the CramerWold device.
= n 1 Lny j
~ y as n ~ oo.
2. (a) The derivative of Fn is fn (x) = 1  cos(2nn x ), for 0 _::: x _::: 1. It is easy to see that fn is
nonnegative and
fn (x) dx = 1. Therefore Fn is a distribution function with density function fn.
(b) As n ~ oo,
sin(2nn x)
1
<~o
2nn
 2nn
'
JJ
247
[5.9.3][5.9.6]
Solutions
and so Fn (x) ~ x for 0 ::= x ::= 1. On the other hand, cos(2nn x) does not converge unless x E {0, 1},
and therefore fn(x) does not converge on (0, 1).
implying that Z
= 2Np
pe't
1  qeit
pe2pit
r/Jz(t) = JN(2pt) = { 1  (1  p)e2pit
4.
{ . }k
r(
}k {
=
}k ~
k
(1  2zt)
All you need to know is the fact, easily proved, that 1/fm (t) = eitm satisfies
1rr:rr 1/fj(t)l/fk(t) d t= {
2n
if j + k = 0,
ifj+k#O,
1:rr ettkifJ(t)dt
.
1 L
=:rr
2n .
00
II"(X = j)
J=00
A.
II"(X = kA) = 2n
Let X be uniformly distributed on [a, a], Y be uniformly distributed on [ b, b ], and let X and
Y be independent. Then X has characteristic function sin(at)j(at), andY has characteristic function
sin(bt)/(bt). We apply the inversion theorem (5.9.1) to the characteristic function of X+ Y to find
that
00
00 sin(at) sin(bt)
1
1
a 1\ b
2
ifJx+y(t) dt = dt = fx+y(O) = b.
2
n oo
2n oo
abt
2a
5.
6.
It is elementary that
248
Solutions
[5.9. 7][5.10.1]
7. The vector X has joint characteristic function (t) = exp( itVt'). By the multidimensional
version of the inversion theorem (5.9.1), the joint density function of X is
j(x) =  1 
(2n)n }ftl.n
Therefore, if i
i=
j,
i=
a
IP'(maxXk:::::
u) =
OVij
j,
 of dx
Q OVjj
J'
where
dx' is an integral over the variables Xk for k i= i, j.
Therefore, IP'(maxk X k ::::: u) increases in every parameter Vij, and is therefore greater than its
value when Vij = 0 fori i= j, namely ITk IP'(Xk::::: u).
8.
By a twodimensional version of the inversion theorem (5.9.1) applied to E(eitX' ), t = (t1, t2),
~ IP'(X 1 >
op
0, X2 > 0) =
("' ("' {
op lo lo
a 1 {{
= op 4n 2 llJRz
=  12
4n
1~
JR2
~
4n
1}JRz
exp(itVt')
dt
(it 1 )(it2 )
1
2Jl' .JjV=Tf =
1
exp( 2J tVt)dt=
~
2
4n
2n y 1  p2
We integrate with respect top to find that, in agreement with Exercise (4.7.5),
IP'(X1 > 0, X2 > 0) =
4 + 2n
sin p.
i).
249
i.
[5.10.2][5.10.4]
Solutions
(b) Let {Xi : i :=:: 1} be a collection of independent Poisson random variables, each with parameter 1.
Then Sn =
Xi is Poisson with parameter n, and by the central limit theorem
'J:'i
nk
(ISnnl
k! = IP'
Jn ~ x ) ~ <l>(x) <1>(x),
as above.
k:
lknl::=:xy'n
2. A superficially plausible argument asserts that, if all babies look the same, then the number X of
correct answers inn trials is a random variable with the bin(n,
distribution. Then, for large n,
i)
X ln
)
IP' ( ~2  > 3 ::::::1 <1>(3)::::::
zJn
rJrm
X ~n
=
~Jn
910 750
~s
5v'15 
Now we might say that the event {X ~n > iJn} is sufficiently unlikely that its occurrence casts
doubt on the original supposition that babies look the same.
A statistician would level a good many objections at drawing such a clear cut decision from such
murky data, but this is beyond our scope to elaborate.
3.
Clearly
l/Jy(t) = E{E(eitY
~t
1  (e 1
1)
)s
1 .
= (1
2 e1
)s
It follows that
1 I
E(Y) = :l/Jy(O) = s,
l
whence var(Y) = 2s. Therefore the characteristic function of the normalized variable Z = (Y EY)/Jvar(Y) is
Now,
1og{ifJy(t/5s)} = s log(2 eitf,ffs) = s(eitj,ffs 1)
=
+ ~s(eitj,ffs 1) 2 + o(l)
it
4. Since X1 is nonarithmetic, there exist integers nJ, n2, ... , nk with greatest common divisor 1
and such that IP'(X 1 = ni) > 0 for 1 ~ i ~ k. There exists N such that, for all n :=:: N, there exist nonnegative integers a1, a2, ... , ak such that n = a1n1 + +aknk. If xis a nonnegative integer, write
250
Solutions
[5.10.5][5.10.5]
N = f31 n 1 + + f3knk> N + x = Y1 n 1 + + Yknk for nonnegative integers f31, ... , f3b Y1 ... , Yk
Now Sn = X1 + + Xn is such that
k
i:::: k)
+ {3;,
+ x)IP'(Sn =
+ x)
> 0 where
N) > 0
where SG,B+G = ~f=+J'+ 1 X;. Also, IP'(Sn Sn,B+G = x) > 0 as required.
5.
Let X 1, X 2, ... be independent integervalued random variables with mean 0, variance 1, span 1,
I 2
and common characteristic function cf>. We are required to prove that JnlP'(Un = x)+ e zx
as n + oo where
1
X1 + X2 + + Xn
Un = JnSn =
Jn
;...fiJi
and xis any number of the form kf Jn for integral k. The case of generaltt and a 2 is easily derived
from this.
By the result of Exercise (5.9.4), for any such x,
IP'(Un
= x) =
r.;;
2n v n
since Un is arithmetic. Arguing as in the proof of the local limit theorem (6),
Now In = 2...fiii ( 1  <I> (n Jn)) + 0 as n + oo, where <I> is the N (0, 1) distribution function. As
for In, pick 8 E (0, n). Then
I 2
The final term involving e zt is dealt with as was ln. By Exercise (5.7.5a), there exists).
such that lc/>(t)l <).if 8:::: ltl:::: n. This implies that
251
(0, 1)
[5.10.6][5.10.7]
Solutions
The proof of this is considerably simpler if we make the extra (though unnecessary) assumption
that m3 =lEI
I < oo, and we assume this henceforth. It is a consequence of Taylor's theorem (see
XI
3
e(t)
Let K 0 = sup{l&(u)l : lui :S 8}, noting that K 0 < oo, and pick 8 sufficiently small that 0 < 8 < rr
and 8K0 <
For It I < 8../fi,
!
an e _lr21
1t1
2
1 2
1t1 _I 12
IcfJ(tfvn;
2
:SK0 .fiiexp(t8K0  2 t):SK0 .fiie 4 ,
3
and therefore
as n+ oo
as required.
6.
lo0e
x2
2x (log x ) 2
dx
~1 e2u
du
oo u 2
(substitute x = eu), a finite integral. Therefore the X's have finite mean and variance. The density
function is symmetric about 0, and so the mean is 0.
By the convolution formula, if 0 < x < e 1,
el
h(x) =
f(y)f(x y)dy
e1
r
Jo
f(y)f(x y)dy
f(x)
r
Jo
f(y)dy,
(x) >
....l.!:!)_ =
 2log lx I
1
41x I(log lx 1) 3
~'
Jn
kn
lxl(log lxl)n+ 1 '
0 < x < e ,
for some positive constant kn. Therefore fn (x) + oo as x + 0, and in particular the density function
of (X 1 + + X n) / .fii does not converge to the appropriate normal density at the origin.
7.
cp(is) =
v2rr
=
1
~
v2rr
lo
=exp(../2i),
252
Solutions
Large deviations
[5.10.8][5.11.3]
by the result of Problem (5.12.18c), or by consulting a table of integrals. The required conclusion
follows by analytic continuation in the upper halfplane. See Moran 1968, p. 27l.
(a) The sum Sn = I;~=l Xr has characteristic function E(eitSn) = rp(t)n = (tn 2 ), whence
Un = Sn/n has characteristic function rp(tn) = E(eitnX I). Therefore,
8.
(b) E(eitTn)
~) ~ 0
asn
oo.
= rp(t) = E(eitX 1 ).
9. (a) Yes, because Xn is the sum of independent identically distributed random variables with
nonzero variance.
(b) It cannot in general obey what we have called the central limit theorem, because var(Xn) =
(n 2  n) var(8) + nE(8)(1  E(8)) and n var(Xl) = nE(8)(1  E(8)) are different whenever
var(8) =F 0. Indeed the right 'normalization' involves dividing by n rather than Jn. It may be shown
when var( e) =F 0 that the distribution of X n / n converges to that of the random variable 8.
2. (i) Let Yn have the binomial distribution with parameters n and i. Then 2Yn  n has the same
distribution as the random variable Sn in Exercise (5.11.1). Therefore, if 0 < a < 1,
(ii) This time let Sn = X 1+ + Xn, the sum of independent Poisson variables with parameter 1. Then
Tn = en!P'(Sn > n(l +a)). The moment generating function of X1  l is M(t) = exp(e 1  1  t),
and the large deviation theorem gives that r,f!n ~ einf1 >o{g(t)} where g(t) =eat M(t). Now
g' (t) = (e 1 a  l) exp(e 1 at  t  1) whence g has a minimum at t = log(a + l). Therefore
r,!ln ~ eg(log(1 +a)) = {ej(a + 1)}a+l.
3. SupposethatM(t) = E(e 1x) isfiniteontheinterval [8, 8]. Now, fora> 0, M(8) 2: e 8a!P'(X >
a), sothat!P'(X >a)_:::: M(8)e 8a. Similarly, IP'(X <a)_:::: M(8)e8a_
J[O,oo)
eltlx dF(x)
[5.11.4][5.12.2]
Solutions
(the term '1' takes care of possible atoms at 0). However 1  F(x) :::: ~teA.x, so that M(t) < oo if
ltl is sufficiently small.
4.
The characteristic function of Sn/n is {elt/nl}n = eltl, and hence Sn/n is Cauchy. Hence
lP'(Sn >an)=
dx
00
n(l
+ x 2)
(n
1 tan 1 a ) .
=n 2
1 6
{6
~s
i}
10
1 ) 10 { 1 _ s6}
6s
~
10
1 ) 10
6
(1  lOs + .. )(1 +lOs + ... ).
6s
The coefficient of s 27 is
2. (a) The initial sequences T, HT, HHT, HHH induce a partition of the sample space. By conditioning
on this initial sequence, we obtain f(k) = qf(k  1) + pqf(k  2) + p 2qf(k  3) fork > 3,
where p + q = 1. Also f ( 1) = f (2) = 0, f (3) = p 3 . In principle, this difference equation
may be solved in the usual way (see Appendix 1). An alternative is to use generating functions.
Set G(s) = 2.::~ 1 sk f(k), multiply throughout the difference equation by sk and sum, to find that
G(s) = p 3s3/{1 qs pqs 2  p 2qs 3 }. To find the coefficient of sk, factorize the denominator,
expand in partial fractions, and use the binomial series.
Another equation for f(k) is obtained by observing that X =kif and only if X > k 4 and the
last four tosses were THHH. Hence
k4
k > 3.
t=l
Applying the first argumentto the mean, we find that f1, = E(X) satisfies f1, = q(1 + f1,) + pq(2+
~t) + p 2q(3 + ~t) + 3p3 and hence 1t = (1 + p + p2)jp3.
As for HTH, consider the event that HTH does not occur in n tosses, and in addition the next
three tosses give HTH. The number Y until the first occurrence of HTH satisfies
lP'(Y > n)p 2q = lP'(Y = n + 1)pq + lP'(Y = n + 3),
:=:::
2.
~{1
+ w 3 + w6 } =
~{1 +w +w 2 } =
~{1+w 2
1,
if k = 3r,
0,
ifk = 3r + 1,
+w4 }=0,
ifk=3r+2,
254
Solutions
Problems
~{GN(l)
2::;!~J JP(N =
[5.12.3][5.12.6]
3. We have that T = k if no run of n heads appears in the first k  n  1 throws, then there is a
tail, and then a run of n heads. Therefore IP(T = k) = JP(T > k n 1)qpn fork :=::: n + 1 where
p + q = 1. Finally JP(T = n) = pn. Multiply by sk and sum to obtain a formula for the probability
generating function G of T:
00
00
G(s)pnsn=qpn L
sk L
JP(T=j)=qpnLIP(T=j) L
sk
k=n+1 j>kn1
j=1
k=n+1
qpnsn+1 oo
.
qpnsn+1
1s LIP(T=j)(lsl)= 1s (1G(s)).
j=1
Therefore
4.
~ sk
L...
(k l)
k=r
r 1
pr (1 p)kr =
(___!!!_)r
l  qs
where p+q = 1. The mean is G'(l) = rjp and the variance is G"(l) +G'(l) {G'(l)} 2 = rqjp 2 .
5.
Po(2n) ~
{nn+2en.J2]r}2
(pq)n
(4 pq)n
;:;;;; .
v7rn
The generating function Fo(s) for the first return time is given by Fo(s) = 1  Po(s) 1 where
Po(s) = l:n s 2n Po(2n). Therefore the probability of ultimate return is Fo(l) = 1 A  1 where, by
Abel's theorem,
.f
1
1 p=q=2>
A= LPo(2n){ = 00
if p =J:. q.
n
<00
Hence Fo(1) = l if and only if p = ~
6.
(a) Rn = X~
+ Y1' satisfies
Hence Rn = n + Ro = n.
(b) The quick way is to argue as in the solution to Exercise (5.3.4). Let Un = Xn + Yn, Vn = Xn Yn.
Then U and V are simple symmetric random walks, and furthermore they are independent. Therefore
255
[5.12.7][5.12.8]
Solutions
by (5.3.3). Using Stirling's formula, Po(2n) ~ (mr) 1, and therefore L:n Po(2n) = oo, implying
that the chance of eventual return is l.
A longer method is as follows. The walk is at the origin at time 0 if and only if it has taken equal
numbers of leftward and rightward steps, and also equal numbers of upward and downward steps.
Therefore
(2n)!
(l)4n (2n) 2
( l)2n n
Po(2n) = 4
(m!)2{(n m)!}2 = 2
n
Let eij be the probability the walk ever reaches j having started from i. Clearly eao =
ea,a1 ea1,a2 e10, since a passage to 0 from a requires a passage to a  1, then a passage
to a 2, and so on. By homogeneity, eao = (ero)a.
7.
Now
lei
> 1, and
ldl
pe3o
+ qeoo = pef0 + q.
The
d= p+Jp2+4pq.
2p
:=:: 1 if and only if p 2 + 4pq :=:: 9p 2 which is to say that p _::: ~It follows that
When p > ~.we have that d < 1, and it is actually the case that e10
eao
p + Jp2 +4pq)a
2p
= d, and hence
'f p > 3
1
In order to prove this, it suffices to prove that eao < 1 for all large a; this is a minor but necessary
chore. Write Tn = Sn  So = I:i= 1 Xi, where Xi is the value of the ith step. Then
eao
= lP'(Tn
= lP'(nJL
00
.::0
where X is a typical step. Now E(e 1 (~tX)) = 1 + o(t) as t 0, and therefore we may pick t > 0
such that e(t) = e 11LE(e 1(tLX)) < 1. It follows that eao .::0 I;~ 1 etae(t)n which is less than 1
for all large a, as required.
8.
We have that
where p + q = 1. Hence Gx,y(s, t) = G(ps + qt) where G is the probability generating function
of X + Y. Now X and Y are independent, so that
G(ps
Solutions [5.12.9][5.12.11]
Problems
9.
var(Zn+1) = o2(~tn
(52fl,n(l _
fl,n+1)
fl,
n ::=::
0,
+ 1).
10. (a) Since the coin is unbiased, we may assume that each player, having won a round, continues
to back the same face (heads or tails) until losing. The duration D of the game equals kif and only
if k is the first time at which there has been either a run of r  1 heads or a run of r  1 tails; the
probability of this may be evaluated in a routine way. Alternatively, argue as follows. We record S
(for 'same') each time a coin shows the same face as its predecessor, and we record C (for 'change')
otherwise; start with a C. It is easy to see that each symbol in the resulting sequence is independent
of earlier symbols and is equally likely to be S or C. Now D = k if and only if the first run of r  2
S's is completed at time k. It is immediate from the result of Problem (5.12.3) that
( .!st2 (1  .!s)
GD(s)= 2
1 s
+ (~sy 1
+ k 1).
n=1
1 {
G D(s)
r1
Wk(s) =  
1
1
2
) G D(w s)
+ kTG
D(ws) + ~(k
ww 1
1
r2 }
+ ... + w(r2)(k1)
G D (w
s)
It may be seen (as for Problem (5.12.2)) that the coefficient of si in Wk(s) is lP'(D = i) if i is of the
form n(r 1) + (k 1) for some n, and is 0 otherwise. Therefore lP'(Ak wins)= Wk(1).
(c) The pool contains D when it is won. The required mean is therefore
JE(D I Ak wins)=
E(DI
{Ak wms}
lP'(Ak wins)
W 1 (1)
= _k_.
Wk(1)
(d) Using the result of Exercise (5.1.2), the generating function of the sequence lP'(D > k), k
T(s) = (1 G D(s))/(1  s). The required probability is the coefficient of sn in T(s).
::=::
0, is
11. Let Tn be the total number of people in the first n generations. By considering the size Z 1 of the
first generation, we see that
ZJ
Tn = 1 + LTn1(i)
i=1
257
[5.12.12][5.12.14]
Solutions
where Tn1 (1), Tn1 (2), ... are independent random variables, each being distributed as Tn1 Using
the compounding formula (5.1.25), Hn(s) = sG(Hn1 (s)).
IZ
m
= O) = lP'(Zn > N, Zm = 0)
lP'(Zm = 0)
=
lP'(Zm = 0
I Zn
+ r)lP'(Zn
lP'(Zmn = O)N+rlP'(Zn = N
= N
+ r)
+ r)
lP'(Zm = 0)
r= 1
:S
= N
lP'(Zm = 0)
r= 1
lP'(Zm = O)N+ 1
lP'(Zm = O)
00
13. (a) We have that Gw(s) = GN(G(s)) = eJ.(G(s) 1). Also, Gw(s) 11n = eJ.((G(s) 1)/n, the
same probability generating function as Gw but with A replaced by Ajn.
(b) We can suppose that H (0) < 1, since if H (0) = 1 then H (s) = 1 for all s, and we may take A = 0
and G(s) = l. We may suppose also that H(O) > 0. To see this, suppose instead that H(O) = 0
so that H(s) = sr "L.'f':=o sj hj+r for some sequence (hk) and some r 2: 1 such that hr > 0. Find a
positive integer n such that r fn is nonintegral; then H(s) 1/n is not a power series, which contradicts
the assumption that H is infinitely divisible.
Thus we take 0 < H(O) < 1, and so 0 < 1  H(s) < 1 for 0::: s < l. Therefore
log H(s) =log( 1  {1  H(s)l) =A ( 1
+ A(s))
where A = log H(O) and A(s) is a power series with A(O) = 0, A(1)
00
.
"'j=
1 ajsl, we have that
l. Writing A(s)
as n + oo. Now H(s) 11n is a probability generating function, so that each such expression is nonnegative. Therefore aj 2: 0 for all j, implying that A(s) is a probability generating function, as
required.
14. It is clear from the definition of infinite divisibility that a distribution has this property if and only
if, for each n, there exists a characteristic function 1/Jn such that cf>(t) = 1/Jn (t)n for all t.
(a) The characteristic functions in question are
"t
4> (t) =
eJ.(e' 1)
cf>(t) =(A.
A It
)JL.
In these respective cases, the 'nth root' 1/Jn of 4> is the characteristic function of the N(~t/n, a 2 jn),
Poisson (A/n), and r(A, ~t/n) distributions.
258
Solutions [5.12.15][5.12.16]
Problems
(b) Suppose that ifJ is the characteristic function of an infinitely divisible distribution, and let 1/Jn be a
characteristic function such thatifJ(t) = 1/Jn(t)n. Now lifJ(t)l ::0 l for all t, so that
11/ln(t)l = lifJ(t)l1/n+ { l
~ lifJ(t)l
lf lifJ(t)l
=I= 0,
= 0.
For any value oft such that ifJ(t) =I= 0, it is the case that 1/Jn (t) + 1 as n + oo. To see this, suppose
< 2n such that 1/Jn (t) + eie along some subsequence.
instead that there exists satisfying 0 <
Then 1/Jn (t)n does not converge along this subsequence, a contradiction. It follows that
1/J(t)
lim
n+oo
=I= 0,
tf ifJ(t) = 0.
Now J is a characteristic function, so that ifJ(t) =I= 0 on some neighbourhood of the origin. Hence
1/f(t) = 1 on some neighbourhood of the origin, so that 1/J is continuous at the origin. Applying the
continuity theorem (5.9.5), we deduce that 1/J is itself a characteristic function. In particular, 1/J is
continuous, and hence 1/f(t) = 1 for all t, by ( *). We deduce that J (t) =I= 0 for all t.
pnlP'(N = n)
2::~1 pkJP'(N = k)
16. Certainly
1  (Pl + P2)) n
Gx(s) = Gx,Y(s, 1) = ( 1
,
 P2 PIS
Gy(t) = Gx,Y(l, t) =
1 (Pl + P2)
Gx+Y(s) = Gx,Y(s, s) = ( 1 _ (Pl + p 2 )s
)n
+ P2)) n ,
1  P1  P2t
( 1  (P1
giving that X, Y, and X+ Y have distributions similar to the negative binomial distribution. More
specifically,
ll"(X = k) =
C+~ )ci(l1
ll"(X
+y
a)n,
= k) = ( n
= P2!(1
lP'(Y = k) = (n
+~ 
 Pl), Y
259
+~ 
1) yk (1  y )n,
= Pl + P2
1),ak(1 _ ,B)n,
[5.12.17][5.12.19]
Solutions
Now
JE(sx I y = y) =
JE(sx I
{Y=yJ
JP'(Y = y)
where A is the coefficient of tYinG x,Y(s, t) and B is the coefficient of tY in Gy(t)o Therefore
JE(sx
Iy
= y) = ( 1 Pl1pls
=(
1pls
1  Pl ) n+y
1pl
1 PIS
17. As in the previous solution,
=
by the substitution u
= b/ y
u2
b 2 u 2 ) } du
2I(l,ab),
loo
oo a2u2
by the substitution x = y 2
(e) Similarly
by substituting x = y20
260
..jii
du=0
2a
1pl
Solutions [5.12.20][5.12.22]
Problems
x > 0.
J2nx3
(c) We have that
w 2 =
fort > 0. It follows that w 2 has the same distribution as 9V = 9X 2 , and so W 2 has the same
distribution as ~ X 2 . Therefore, using the fact that both X and W are symmetric random variables, W
has the same distribution as
2n
=
1N
1"
1m
N
N+oo
1  eith itx,~.( ) d
e
'~" t
t.
it
Since 14>1 is integrable, we may use the dominated convergence theorem to take the limit as h ..j, 0
within the integral:
f(x)
lim
___!__
2n
N+oo
1N
eitxrf>(t)dt.
N
The condition that 4> be absolutely integrable is stronger than necessary; note that the characteristic
function of the exponential distribution fails this condition, in reflection of the fact that its density
function has a discontinuity at the origin.
21. Let Gn denote the probability generating function of Zn. The (conditional) characteristic function
of Zn/fl.n is
lE(eitZn/JLn
IZ
>
0) =
Gn(eit/JLn) Gn(O).
1  Gn(O)
261
[5.12.23][5.12.24]
Solutions
for all t, if and only if X and X have the same characteristic function, or equivalently the same
distribution.
23. (a) U = X + Y and V = X  Y are independent, so that r/Ju + v = r/Jur/Jv, which is to say that
2X = X+YrfJXY or
Therefore
1/f(t) = 1fr(1t)2 = 1fr(!t)4 = ... = 1/f(t/2n)2n
for n:::: 0.
However, as h + 0,
{1! .._
2
22n
+o(t2/22n)}z2n+ e1t2
for n :::: 1
as n+ oo,
However, by (*),
a2t/
at
=2{"(s)(s)'(s)2},
t=o
= 1.
1 2
262
Solutions [5.12.25][5.12.27]
Problems
=
=
L:
f(x)f(y x)dx
100 {J(x) +
JT(4 + y ) oo
JT(4 + y )
+ J g(y)
where
1= L:{xf(x)+(yx)f(yx)}dx
lim
M,N+oo
M
=0.
Finally,
fz(z) = 2fx+Y(2z) =
1
JT(1
+ z2 )
1_00
2eitx
oo
rjJ(t)
e:n:x
+ e :n:x dx.
where C is a rectangular contour with vertices at K, K + i. The integrand has a simple pole at
1
00
cosh(Jrx)
k=O
263
1)Jrlxl}.
[5.12.28][5.12.30]
(b) Define </J(t)
Solutions
2rr
= 0 otherwise.
1, and </J(t)
Then
1oo
11 eitx (1 It I) dt
oo eitx</J(t) dt = __!__
2rr
1
= 1 lo1 (1 
t) cos(tx) dt
7i
=  1 2 (1 cosx).
liX
1oo
00
eitxexex dx =
lo
L:
i</J'(O)
1
1 it
+ 1 +it
= 1
r(z) =
.
hm
n!nz
n>oo z(z + 1) (z + n)
r , (1) =
hm
n>oo
+ t2.
= 1. By differentiation,
= y.
= JEy'2{1 cos(tX)}
.:S lEitXI
</J(s, t)
= "'
L....t
m<M
n?:_N
smtn
</Jmn(O, 0)
m!n!
264
+ RMN(s, t)
Solutions [5.12.31][5.12.33]
Problems
where ifJmn is the derivative of ifJ in question, and RM N is the remainder. However, subject to appropriate conditions,
{
Jrrl,r1]
L:
3{1cos(tx)}dF(x)
J[rl,r1]
_:::: 3
!
t
r {1cos(vx)}dvdF(x)
x=oo !t Jv=O
= loo (1 sin(tx)) dF(x)
oo
tx
r{1ReJ(v)}dv=1
00
2: {
x:
Jo
Jltxl~1
(1
sin(tx)) dF(x)
tx
since 1  (tx) 1 sin(tx) 2: 0 if !tx! < 1. Also, sin(tx) _:::: (tx) sin 1 for !tx! 2: 1, whence the last
integral is at least
{ x: (1 sin 1) dF(x) 2:
Jltxl~1
~lP'(IXI
2: t 1).
1/JA (t) =IE{ exp(it(X A)/.J'i..)} = exp{ A(eit/v'A 1)  it.J'i..} = exp{ ~t 2 + o(l)}
as A+ oo. Now use the continuity theorem.
(b) In this case,
so that, as A + oo,
it
t2
1 ) +t.
1 2
loglfrA(t)=itv'AAlog ( 1 it ) =itv'A+A ( +o(A)
,ft.
,ft.
265
2A
[5.12.34][5.12.36]
Solutions
z~ n
o)
::=:
+ <f>(O)
where <f> is the N(O, 1) distribution function. The left hand side equals lfD(Zn ::=: n) =
34. If you are in possession of r  1 different types, the waiting time for the acquisition of the next
new type is geometric with probability generating function
Gr(s)
Therefore the characteristic function of Un
= (Tn
II
+ l)s
(n r
n (r l)s
 n log n) / n is
II
.
n
.
. n {(nr+l)eitjn}
nitnl
1/ln(t) = eztlogn
Gr(eztfn) = nzt
=
.
.
n (r l)elt/n
rrnl(neztjn r)
r=l
r=l
r=O
The denominator satisfies
n1
n1
r=O
r=O
n+oo
nitn'
lim
I
.
n+00 rr~,:o(n it r)
= r ( l  it),
as n + oo
+ r(z)
the convergence being uniform on a11y region of the complex plane containing no singularity of r.
The claim now follows by the result of Problem (5.12.27c).
1
n
</Jn(t)
1 .t
 e 1 x dx
n 2n
sin(nt)
nt
1
if t
=f. 0,
if t = 0.
It follows that, as n + oo, <Pn (t) + 8o1 , the Kronecker delta. The limit function is discontinuous at
t = 0 and is therefore not itself a characteristic function.
36. Let G i (s) be the probability generating function of the number shown by the i th die, and suppose
that
12
G1 (s)G2(s) =
so that 1 s 11
However
1 k
{;2 rrs
=
1 s 11
s 2(1 s 11)
11 (1 _ s) ,
266
Solutions [5.12.37][5.12.38]
Problems
where w1, w1, ... , ws, ms 'are the ten complex eleventh roots of unity. The Wk come in conjugate
pairs, and therefore no five of the ten terms in rr~=1 (Wk  S )(Wk  S) have a product Which is a real
polynomial. This is a contradiction.
37. (a) Let H and T be the numbers of heads and tails. The joint probability generating function of
Hand Tis
= GN(qt + ps) =
exp {A.(qt
+ ps 1)}.
It follows that
GT(f)
where M(t) = E(etYt) = (q +pet). With the aid of a little calculus, we find that
lP'(Sn 0:: n/3)1/n +
(~ ) {3(11 =~ )1{J
Hence
E{Xn(f3n)}+ { 0
00
where
y(f3) = 2
(~r
p::S/3<1.
~fy(f3) < 1,
tf y(/3) > 1,
c=
~r{3
is a decreasing function of {3. If p < ~,there is a unique f3c E [p, 1) such that y(f3c) = 1; if p ::::
then y ({3) > 1 for all f3 E [p, 1) so that we may take f3c = 1.
Turning to the final part,
1P'(Xn(f3n) 0::
1)
::S E{Xn(f3n)}+ 0
267
if f3 > f3c.
[5.12.38][5.12.38]
Solutions
for any N taking values in the nonnegative integers. This is easily proved: certainly
var(N
IN
=I= 0) = E(N 2
IN
=I= 0)  E(N
IN
=I= 0) 2 ::: 0,
whence
E(N 2)
E(N) 2
>
.
W'(N =/= 0)  W'(N =I= 0)2
We have that E{Xn (.Bn) 2 } = L:n,p E(Inlp) where the sum is over all such paths n, p, and In is the
indicator function of the event {B(n) ::: ,Bn }. Hence
E{Xn(,8n) 2 } = LE(/n)
+L
E(/nlp)
= E{Xn(,Bn)} + 2n 1 L
n#p
E(hlp)
p#L
where L is the path which always takes the left fork (there are 2n 1 choices for n, and by symmetry
each provides the same contribution to the sum). We divide up the last sum according to the number
of nodes in common top and L, obtaining 2::~:,\ 2nm 1E(hlM) where M is a path having exactly
m nodes in common with L. Now
E(h/M) = E(IM
Ih
where Tnm has the bin(n  m, p) distribution (the 'most value' to IM of the event {h
obtained when all m nodes in L n M are black). However
1} is
E(N2) :S E(N)
+ 2n1
n1
n1 (
2nm1. kE(h)E(/M)
= E(N) + iE(N)2 L
m=1
m=1
21 )
P
whence, by (*),
 E(N)1
+ i 2::~:,\ (2p)m
If ,8 < ,Be then E(N) + oo as n + oo. It is immediately evident that W'(N =I= 0) + 1 if p ::;
0) ::: c(,B)
for all n
where c(,B) is some positive constant. Find E > 0 such that ,8 + E < .Be Fix a positive integer m, and
let :Pm be a collection of 2m disjoint paths each of length n  m starting from depth m in the tree.
Now
P(Xn(,Bn) = 0) :S P( B(v) < ,Bn for all v E 9'm) = P( B(v) < ,Bn ) 2m
where v
9'm . However
P(B(v) < ,Bn) :S P(B(v) <
+ E)(n 
268
(,8
+ E)(n m))
+ E)m/E.
Problems
Solutions
[5.12.39][5.12.42]
by(**); we let n+ oo and m+ oo in that order, to obtain lP'(Xn(.Bn) = 0)+ 0 as n+ oo.
Peitfn .
E(eitYnfn) =
+ _ _
A it
1  (1 p)elt/n
as n + oo, the limit being the characteristic function of the exponential distribution.
40. If you cannot follow the hints, take a look at one or more of the following: Moran 1968 (p. 389),
Breiman 1968 (p. 186), Loeve 1977 (p. 287), Laha and Rohatgi 1979 (p. 288).
41. With Yk = kXk. we have that E(Yk)
Y1 + Y2 + + Yn is such that
0, var(Yk)
k 2 , EIYf I
k 3 . Note that Sn
as n + oo, where cis a positive constant. Applying the central limit theorem ((5.10.5) or Problem
(5.12.40)), we find that
Sn
D
as n + oo,
~ + N(O, 1),
yvarSn
where var Sn
= l:k= 1 k 2 ~
1n 3 as n+ oo.
42. We may suppose that 11 = 0 and a = 1; if this is not so, then replace Xi by Y; = (Xi  11) /a.
Lett = (to, t1, t2, ... , tn) E JRn+l, and set 7 = n  1 l:J= 1 fj. The joint characteristic function of the
fl
exp (
exp
(i [~ +
fj 7]
Xj)}
j=1
]=1
~ [ ~ + fj  7r)
by independence. Hence
ifJ(t)
= exp (  1 Ln
whence
s2 =
[t
_Q
2 j= 1 n
+ (fj
7)
]2) = {
t2
1 L(tj
n
exp _...Q_7) 2 }
2n
2 j= 1
zJ.
269
is independent of
[5.12.43][5.12.47]
Solutions
43. (i) Clearly, lP'(Y ::; y) = lP'(X::; logy)= cp(log y) for y > 0, where cp is the N(O, 1) distribution
function. The density function of Y follows by differentiating.
(ii) We have that fa(x) :=::: 0 if /a/::; 1, and
lo
oo
asin(2nlogx)
00
1
I I
2
1
I 2
M"::"ez(ogx) dx=
M"::"asin(2ny)ezY dy=O
xv2n
oo v2n
since sine is an odd function. Therefore J~oo fa (x) dx = 1, so that each such fa is a density function.
For any positive integer k, the kth moment of fa is J~00 xk f(x) dx + la(k) where
la(k) =
1
k
I 2
M"::"a sin(2ny)e YzY dy = 0
oo v2n
00
since the integrand is an odd function of y k. It follows that each fa has the same moments as f.
44. Here is one way of proving this. Let X 1, X 2, . . . be the steps of the walk, and let Sn be the
position of the walk after the nth step. Suppose tt = E(X ,) satisfies tt < 0, and let em = lP'(Sn =
0 for some n :=::: 1 I So= m) where m > 0. Then em ::; 2:~ 1 1P'(Tn > m) where Tn = x, + X2 +
+ Xn = Sn  So. Now, fort > 0,
lP'(Tn > m)
= lP'(Tn
where M(t) = E(et(XIJJ.)). Now M(t) = 1 +O(t 2) as t + 0, and therefore there exists t (> 0) such
thate(t) = e 1 f.J.M(t) < 1 (rememberthattt < 0). Withthischoiceoft, em::; 2::~ 1 etme(t)n+ 0
as m + oo, whence there exists K such that em < for m :=::: K.
Finally, there exist 8, E > 0 such that lP'(X, < 8) > E, implying that lP'(SN < K I So= 0) >
EN where N = K I 8l, and therefore
:=:::
therefore the walk is transient. This proof may be shortened by using the BorelCantelli lemma.
45. Obviously,
L={~,+L
ifX,>a,
ifX1 ::;a,
47. Let Wr be the number of flips until you first see r consecutive heads, so that lP'(Ln < r) =
lP'(Wr > n). Hence,
oo
oo
1 E(s Wr)
1 + """snlP'(Ln < r) = """snlP'(W > n) =    L.J
L.J
1s
n=l
n=O
270
Solutions
Problems
[5.12.48][5.12.52]
{ iXn
I
zXn
with probability
i,
with probability
i.
+ Yn
itX
I
I
I
I
I
A
n+l) = zn(zt) + zn(zt).
A zt
,
A
I .
zlt
A
4!(
n
A!
A
lf
2n
.t
!
+ ,;t
A!
(a) (1eA.)/A, (b) (pjq 2 )(q+1og p), (c) (1qn+l )/[(n+ 1) p ], (d) [l+(p/q) log p]jlog p.
(e) Not if lP'(X + 1 > 0) = 1, by Jensen's inequality (see Exercise (5.6.1)) and the strict concavity of
the function f(x) = 1jx. If X+ 1 is permitted to be negative, consider the case when lP'(X + 1 =
1) =lP'(X+ 1 = 1) =
Px(t)
 _AP_
1 qx(t)
Ap it'
51. Consider the function G(x) = {E(X 2 )} 1 J~oo y 2 dF(y). This function is rightcontinuous and
increases from 0 to 1, and is therefore a distribution function. Its characteristic function is
1
1
i, lxl <
1,
f(x,zx)dx=
,\(z + 2)
1
4 (2z)
1
f(x,
y) i=
if  2 <
fx(x)fy(y), X andY
z<
0,
if0<z<2,
the 'triangular' density function on ( 2, 2). This is the density function of the sum of two independent
random variables uniform on (1, 1).
271
6
Markov chains
I X1
whence the sequence is a Markov chain. The chain is homogeneous if the Xi are identically distributed.
2.
(a) With Yn the outcome of the nth throw, Xn+l = max{Xn, Yn+d, so that
0
{ I .
Pij =
if j < i
=i
if j
if j > i,
ifj<i
'f.
.
1 1 = !.
If j > i, then Pij(n) = lP'(Zn = j), where Zn = max{Y1, Y2, ... , Yn}, and an elementary calculation
yields
Pi} (n) =
i < j:::: 6.
(b) Nn+l Nn is independent of N1, N2, ... , Nn, so that N is Markovian with
i
{
Pi} =
t
if j = i
if j
+ 1,
= i,
otherwise.
Cr+l = {
~r + 1
Pi}=
j =0,
j = i
otherwise.
272
+ 1,
Markov processes
Solutions
[6.1.3][6.1.4]
Br 1
if Br > 0,
if Br = 0,
Yr
(65)j1_61
3.
(*)
lP'(Xn+! = i
+ 1 I Xn
if j = i  1 ~ 0,
.f.
0 . 1
ll=,j~.
1,
= i, B) = lP'(Xn+! = i + 1 I Sn = i, B)lP'(Sn = i I Xn = i, B)
+ lP'(Xn+! = i + 1 I Sn = i, B)lP'(Sn = i I Xn = i, B)
+ 1 I Sn =
i, B)= p,
lP'(Xn+! = i
+ 1 I Sn
= i, B)= q,
where p (= 1  q) is the chance of a rightward step. Let l be the time of the last visit to 0 prior to
the time n, l = max{r : ir = 0}. During the timeinterval (1, n], the path lies entirely in either the
positive integers or the negative integers. If the former, it is required to follow the route prescribed by
the event B n {Sn = i }, and if the latter by the event B n {Sn = i}. The absolute probabilities of
these two routes are
whence
lP'(Sn = i
n, +
I Xn = i, B) =
Jl"i
7r2
L
= 1  lP'(Sn = i I Xn = i, B).
pz + qz
.
.
pi+!+ qi+!
lP'(Xn+! = l + 1 I Xn = l, B)=
.
.
= 1 lP'(Xn+l = i  1 I Xn = i, B).
pl +ql
Finally lP'(Xn+l = 1 I Xn = 0, B) = 1.
(ii) If Yn > 0, then Yn  Yn+! equals the (n + l)th step, a random variable which is independent of
the past history of the process. If Yn = 0 then Sn = Mn, so that Yn+! takes the values 0 and 1 with
respective probabilities p and q, independently of the past history. Therefore Y is a Markov chain
with transition probabilities
fori > 0,
Pij =
ifj=i1
'f .
.+1
1 J = l
'
POj = { q
ifj=O
ifj=l.
4.
I Yr = ir for 0
~ r ~
k) =
=
=
ns)
k!
273
I Yk = ik),
[6.1.5][6.1.9]
Markov chains
Solutions
if j = i
{ 2pq
q2
if j
+ 2,
= i,
if j = i 2.
(b) With the usual notation, the transition probability lrij is the coefficient of sj in G(G(s))i.
5.
.)
( I
lP' F /(X)= 1, Xn = z =
lP'(F,/(X)=l.Xn=i)
(
.)
lP' /(X)= 1,Xn = I
where F is any event defined in terms of Xn, Xn+1 .... Let A be the set of all sequences x =
(x1, x2, ... , Xn1, i) of states such that I (x) = 1. Then
lP'(F, /(X)= 1, Xn =
i)
XEA
by the Markov property. Divide through by the final summation to obtain lP'(F /(X) = 1, Xn =
i) = lP'(F I Xn = i).
6.
Let Hn = {Xk = Xk for 0:::: k < n, Xn = i}. The required probability may be written as
lP'(XT+m = j, HT)
lP'(HT)
lP'(HT)
Now lP'(XT+m = j I HT, T = n) = lP'(Xn+m = j I Hn, T = n). Let I be the indicator function of
the event Hn n {T = n }, an event which depends only upon the values of X 1, X2, ... , Xn. Using the
result of Exercise (6.1.5),
lP'(Xn+m = j
I Hn,
T = n) = lP'(Xn+m = j
I Xn
= i) = Pij(m).
Hence
lP'(X
7.
T +m  1
lP'(HT)
 PJ m
Clearly
lP'(Yn+1 = j
I Yr
I Xr
It is easy to find an example in which h is not oneone, for which X is a Markov chain but Y is
not. The first part of Exercise (6.1.3) describes such a case if So =f::. 0.
8.
Not necessarily! Take as example the chains SandY of Exercise (6.1.3). The sum is Sn
+ Yn
I Xm
274
I Xm+r1
= im+r1).
Solutions [6.1.10][6.2.1]
Classification of states
(b) Let {even}= {X2r = i2r forO:::: r:::: m} and {odd}= {X2r+1 = i2r+1 forO :S r :S m 1}.
Then,
I
(k, l) I Yn =
IP'(Yn+1 = (k, l) Yo = (io, i1), ... , Yn = Cin, k)) = IP'(Yn+1 = (k, l) Xn+1 = k)
= IP'(Yn+1 =
Cin, k)),
lP'(Yn+1 = k I Yi =Xi
+ Xi1 for 1 :S i
:S n) = lP'(Yn+1 = k I Xn = Xn)
12. With 1 a row vector of 1's, a matrix Pis stochastic (respectively, doubly stochastic, substochastic)
ifP1' = 1 (respectively, 1P = 1, P1' :::: 1, with inequalities interpreted coordinatewise). By recursion,
P satisfies any of these equations if and only if pn satisfies the same equation.
[6.2.2][6.3.1]
Solutions
Markov chains
2. Let i (# s) be a state of the chain, and define ni = min{n: Pis(n) > 0}. If Xo = i and Xn; = s
then, with probability one, X makes no visit to i during the intervening period [1, n; 1]; this follows
from the minimality of n;. Now s is absorbing, and hence
lP'(no return to i
I Xo = i)
;::: lP'(Xn;
= s I Xo = i)
> 0.
3. Let h be the indicator function of the event {Xk = i}, so that N = I:~o
visits to i. Then
00
h is the number of
00
k=O
which diverges if and only if i is persistent. There is another argument which we shall encounter in
some detail when solving Problem (6.15.5).
4. We write lP'iO = lP'( I Xo = i). One way is as follows, another is via the calculation of Problem
(6.15.5). Note that lP';(Yj ;::: 1) = lP';(1j < oo).
(a) We have that
by the strong Markov property (Exercise (6.1.6)) applied at the stopping time Ti. By iteration, lP'i (Vi ;:::
n) = lP'i ( V; ;::: l)n, and allowing n + oo gives the result.
(b) Suppose i = j. Form ;::: 1,
by the strong Markov property. Now let m + oo, and use the result of (a).
5. Let e = lP'(1j < Ti
before visiting i. Then
Xo
= i) = lP'(Ti
lP'(N:::: 1 I Xo
Likewise, lP'(N:::: k I Xo
< 1j
Xo
= i) = lP'(1j
< T;
I Xo = i)
=e.
1, whence
00
E(N 1 Xo
= i) = "Leo  e)kl = 1.
k=l
276
Solutions
Classification of chains
[6.3.2][6.3.3]
It is easy to see that JE:(Tp)) = 1 + (i  1)/(1  r) if i 0:: 1, since the waiting time at each
intermediate point has mean (1 r) 1. The number N of such 'small' excursions has mass function
f'(N = n) = ai(l ai)n, n :=: 0, where ai = L~i aj; hence JE:(N) = (1 ai)/ai. Each such small
excursion has mean duration
i1 (
Eo
 1
1r
+1
__!!1_ 1 +
i1
1ai
Eo
1{
= (1  ai)
ai
i1
+L
Jaj
(1ai)(lr)
and therefore
E(T?))
Jaj
. 1 r
]=0
By a similar argument,
E(T?))
1
=""'
00
a~
j=i
1 + J  t' ) aj.
1r
and a similar argument yields JE:(To) = 1 + Lj jaj j(l r). The apparent simplicity of these formulae
suggests the possibility of an easier derivation; see Exercise (6.4.2). Clearly JE:(Ti) < oo for i ::::: J
whenever Lj jaj < oo, a condition which certainly holds if J < oo.
2. Assume that 0 < p < 1. The mean jumpsize is 3p  1, whence the chain is persistent if and
only if p =
see Theorem (5.10.17).
1;
3. (a) All states are absorbing if p = 0. Assume henceforth that p =f. 0. Diagonalize P to obtain
P = BAB 1 where
B~ (:
~}
1
0
1
P"
~ BA'n
B1=
0
~ 0 gp)'
0
l2p
0
A=
Therefore
(! \).
I
(I
2
0
) B 1
0 ) .
l4p
(1 4p)n
[6.3.4][6.3.5]
Markov chains
Solutions
Pn (s)
p 33 (s)
Pzz(s) =
1
4(1  s)
+ 2{1 
1
s(l 2p)}
1
2(1  s)
+ 2{1 
1
.
s(l 4p)}
+ 4{1 
1
s(l 4p)}'
Mter a little work one obtains the mean recurrence times Jti = F{i (1): ttl = tt3 = 4, tt2 = 2.
(b) The chain has period 2 (if p =f. 0), and all states are nonnull and persistent. By symmetry, the
mean recurrence times Jti are equal. One way of calculating their common value (we shall encounter
an easier way in Section 6.4) is to observe that the sequence of visits to any given state j is a renewal
process (see Example (5.2.15)). Suppose for simplicity that p =f. 0. The times between successive
visits to j must be even, and therefore we work on a new timescale in which one new unit equals two
old units. Using the renewal theorem (5.2.24), we obtain
2
Pij(2n
+ 1)+
 i f Jj1tj
il is odd;
note that the mean recurrence time of j in the new timescale is ittj. Now "E,j Pij (m) = 1 for all m,
and so, letting m = 2n + oo, we find that 4/ tt = 1 where tt is a typical mean recurrence time.
There is insufficient space here to calculate Pij (n). One way is to diagonalize the transition
matrix. Another is to write down a family of difference equations of the form PI2 (n) = p P22 (n 1) + (1  p) P4z(n 1), and solve them.
4. (a) By symmetry, all states have the same meanrecurrence time. Using the renewalprocess
argument of the last solution, the common value equals 8, being the number of vertices of the cube.
Hence ttv = 8.
Alternatively, lets be a neighbour of v, and lett be a neighbour of s other than v. In the obvious
notation, by symmetry,
 1 + 4/tsv.
3
ltvI
1ttv = 1 + 21tsv
I
I
+ 4/ttv
+ 41twv,
1tsv = 1 +
1twv = 1 +
I
+ 'J.Ittv.
I
3
41twv + 4/ttv,
4/tsv
(c) The required number X satisfies lP'(X = n) = en! (1  e) 2 for n ?: 1, where e is the probability
that the first return of the walk to its starting point precedes its first visit to the diametrically opposed
vertex. Therefore
00
JE(X)
I>en1(1 e)2 = 1.
n=I
5.
I Xo =
+ lP'i(V;
= oo),
by the strong Markov property. Since i + j, we have that lP'j (Vi = oo) ?: 1, which implies IJji = 1.
Also, lP'i (1j < oo) = 1, and hence j + i and j is persistent. This implies IJij = 1.
(b) This is an immediate consequence of Exercise (6.2.4b).
278
Classification of chains
Solutions
x1
= k) =
For j
E A.
[6.3.6][ 6.3.9]
2.:: PjkTJk
k
kES
If x = (Xj : j E S) is any nonnegative solution of these equations, then Xj = 1 ::: T/j for j E A. For
j rf A,
Xj
=L
kES
PjkXk = L Pjk
kEA
= IP'j(TA = 1)
+L
1)
+L
PjkXk
krjA
L Pjk{ L Pki
krjA
iEA
krjA
+L
irjA
+L
krjA
Pjk L PkiXi.
irjA
rf. A,
rf.
7. The first part follows as in Exercise (6.3.6). Suppose x = (xj : j E S) is a nonnegative solution
to the equations. As above, for j rf A,
Xj = 1 + LPjkXk = IP'j(TA::: 1)
+L
= IP'j(TA ::: 1)
Pjk(1
+ LPkiXi)
~A
+ IP'j(TA
::: 2)
+ + IP'j(TA
i~
::: n)
+L
where the penultimate sum is over all paths of length n that do not visit A. We let n + oo to obtain
thatxj :::Ej(TA)=Pj
8. Yes, because the Sr and Tr are stopping times whenever they are finite. Whether or not the exit
times are stopping times depends on their exact definition. The times Ur = min{k > Uri : Xu, E
A, Xu,+! rf. A} are not stopping times, but the times Ur + 1 are stopping times.
9. (a) Using the aperiodicity of j, there exist integers TJ, r2, ... , rs having highest common factor
1 and such that Pjj (rk) > 0 for 1 ::::: k ::::: s. There exists a positive integer M such that, if r ::: M, then
r = Z::~=l akrk for some sequence aJ, a2, ... , as of nonnegative integers. Now, by the ChapmanKolmogorov equations,
s
+ m)::: Pij(m)Pjj(r)
> 0
ifr::: M.
(b) Since there are only finitely many pairs i, j, the maximum R(P) = max{N(i, j) : i, j E S} is
finite. Now R(P) depends only on the positions of the nonnegative entries in the transition matrix P.
279
[6.3.10][6.3.10]
Solutions
Markov chains
There are only finitely many subsets of entries of P, and so there exists f(n) such that R(P) ::": f(n)
for all relevant n x n transition matrices P.
(c) Consider the two chains with diagrams in the figure beneath. In the case on the left, we have that
Pll (5) = 0, and in the case on the right, we may apply the postage stamp lemma with a = n and
b=n1.
3
2
n
10. Let Xn be the number of green balls after n steps. Let ej be the probability that Xn is ever zero
when Xo = j. By conditioning on the first removal,
j+2
j
ej = 2 (j + 1) ej+1 + 2 (j + 1) ej1
j::: 1,
(*)
where
j+2
Pj = 2(j + 1)'
"'
f::QP1P2Pj1
= 2   ~ 2
j+1
as J. ~ oo.
By the result of Exercise (6.3.6), we seek the minimal nonnegative solution (ej) to(*), which is
attained when 2(1  q) = 1, that is, e1 =
Hence
j1
ej =
1! L
q1q2 .. qj1
2 r=O P1P2 Pj1
+1
For the second part, let dj be the expected time until j  1 green balls remain, having started with j
green balls and j + 2 red. We condition as above to obtain
j
dj = 1 + 2(j + 1) {dj+1 +dj}.
We set ej = dj  (2j + 1) to find that (j + 2)ej = jej+b whence ej = ij (j + 1)e1. The expected
time to remove all the green balls is
n
n
n
dj =
ej + 2(j  1)} = n(n + 2) + e1
j (j + 1).
L{
L!
j=1
j=1
j=1
The minimal nonnegative solution is found by setting q = 0, and the conclusion follows by Exercise
(6.3.7).
280
Solutions
[6.4.1][6.4.2]
oo
as n + oo, assuming q < 1. This infinite product is therefore the probability generating function of
the stationary distribution whenever this exists. If G(s) = el.(s1), then
IT
r=O
r=O
Hence
sG(s) = nosA(s) + rs(G(s) no)+ (1 r)(G(s) no)
where A (s) =
A'(1)+1r)
G(1) =no (
.
1r
There exists a stationary distribution if and only if r < 1 and A' (1) < oo, in which case
sA(s) (1  r + sr)
G(s)        (s 1)(A'(l) + 1 r)
Hence the chain is nonnull persistent if and only if r < 1 and A' (1) < oo. The mean recurrence time
j;.,i is found by expanding G and setting fLi = 1/ni.
(6.3.2): Assume that 0 < p < 1, and suppose first that p =f. ~ Look for a solution {Yj : j =f. 0} of
the equations
Yi =
L PijYj.
j10
281
i =f. 0,
[6.4.3][6.4.3]
Solutions
Markov chains
as in (6.4.10). Away from the origin, this equation is Yi = qyi1 + PYi+2 where p + q = 1, with
auxiliary equation pe 3  e + q = 0. Now pe 3  e + q = p(e  1)(8  a)(8  /3) where
a=
if i :::: 1,
if i :::; 1,
the constants A, B, C, D being chosen in such a manner as to 'patch over' the omission of 0 in the
equations (*):
Y2 = qy3,
Y1 = PY3
The result is a bounded nonzero solution {Yj} to (*), and it follows that the chain is transient.
For p < ~, follow the same route with
if i :::: 1,
ifi:::;1,
the constants being chosen such that Y2 = qy_3, Y1 = qY2
Finally suppose that p = ~,so that a= 2 and f3 = 1. The general solution to(*) is
Yi = { A+ Bi + Cai.
D+ Ei +Fa 1
if i :::: 1,
ifi::::: 1,
subject to (** ). Any bounded solution has B = E = C = 0, and (**) implies that A = D = F = 0.
Therefore the only bounded solution to(*) is the zero solution, whence the chain is persistent. The
equation x = xP is satisfied by the vector x of 1's; by an appeal to (6.4.6), the walk is null.
(!, 1, !)
The quantities X 1, X2, ... , Xn depend only on the initial contents of the reservoir and the rainfalls
Yo, Y1, ... , Yn1 The contents on day n + 1 depend only on the value Xn of the previous contents
and the rainfall Yn. Since Yn is independent of all earlier rainfalls, the process X is a Markov chain.
Its state space isS= {0, 1, 2, ... , K 1} and it has transition matrix
[go+
lP'=
go
0
Kl
g2
g1
go
g3
g2
g1
gK1
gK2
gK3
GK1
Gx
GK2
go
G1
282
0 < r < K 1,
Solutions
[6.4.4][6.4.5]
The final equation is a consequence of the previous ones, since L;~(/ Jri = 1. Suppose then that
v = ( v1, v2, ... ) is an infinite vector satisfying
Multiply through the equation for Vr by sr+ 1, and sum over r to find (after a little work) that
00
N(s) =
00
G(s) = Lgisi
VjSi,
i=O
i=O
The probability generating function of the Jri is therefore a constant multiplied by the coefficients of
s0 , s 1, ... , sK  1 in go(s 1)/(s G(s)), the constant being chosen in such a way that L;~0 1 rri = 1.
p(l  qs)
sG(s)
pqs
==' =
= p
q
+ _ ______;:...__
1(qsjp)
The coefficient of s0 is 1, and of si is qi+ 1j pi if i ::: 1. The stationary distribution is therefore given
byrri =qno(qfpi fori::: 1,where
+ 1) if p =
if p
4.
q =
P,~ (~
1
2
1
2
0
0
0
0
1
2
1
2
i i p, i (1 
p),
i(1 
P))
5. (a) Set i = 1, and find an increasing sequence n1 (1), n1 (2), ... along which x1 (n) converges.
Now set i = 2, and find a subsequence of (n1(j) : j ::: 1) along which x2(n) converges; denote
this subsequence by n2(1), n2(2), .... Continue inductively to obtain, for each i, a sequence Di =
(ni (j) : j ::: 1), noting that:
(i) Di+1 is a subsequence ofni, and
(ii) limr+ooXi(ni(r)) exists for all i.
Finally, define m k = n k ( k). For each i ::: 1, the sequence m i, m i+1, . . . is a subsequence of Di , and
therefore limr+oo Xi (mr) exists.
(b) Let S be the state space of the irreducible Markov chain X. There are countably many pairs i, j
of states, and part (a) may be applied to show that there exists a sequence (nr : r ::: 1) and a family
(aij : i, j E S), not all zero, such that Pij (nr) + Ciij as r + oo.
283
[6.4.6][6.4.10]
Solutions
Markov chains
Now X is persistent, since otherwise Pi} (n) + 0 for all i, j. The coupling argument in the proof
of the ergodic theorem (6.4.17) is valid, so that PaJ (n)  Pbj (n) + 0 as n + oo, implying that
aaj = abj for all a, b, j.
6.
7.
Let Xn be the Markov chain which takes the valuer if the walk is at any of theY nodes at level
r. Then Xn executes a simple random walk with retaining barrier having p = 1  q = ~,and it is
thus transient by Example (6.4.15).
8.
Assume that Xn includes particles present just after the entry of the fresh batch Yn. We may write
Xn
Xn+l =
L Bi,n + Yn
i=l
where the Bi,n are independent Bernoulli variables with parameter 1  p. Therefore X is a Markov
chain. It follows also that
= 1 (~)n + (~)nXo.
The equilibrium probability generating function satisfies
whence
d
ds {0s)G(s)} = sG(s),
subject to G(1) = 1. The solution is G(s) = esl, which is the probability generating function of
the Poisson distribution with parameter 1.
10. This is the claim of Theorem (6.4.13). Without loss of generality we may takes = 0 and the Yj to
be nonnegative (since if the Yj solve the equations, then so do Yj + c for any constant c). LetT be the
matrix obtained from P by deleting the row and column labelled 0, and write Tn = (tij (n) : i, j # 0).
Then Tn includes all the nstep probabilities of paths that never visit zero.
We claim first that, for all i, j it is the case that tij (n) + 0 as n + oo. The quantity tij (n) may
be thought of as the nstep transition probability from i to j in an altered chain in which s has been
made absorbing. Since the original chain is assumed irreducible, all states communicate with s, and
therefore all states other than s are transient in the altered chain, implying by the summability of fij (n)
(Corollary (6.2.4)) that tij (n) + 0 as required.
Iterating the inequality y :=: Ty yields y :=: Tny, which is to say that
00
00
284
i ::: 1.
Solutions [6.4.11][6.4.12]
1,
00
lP'(Aoo
I Xo = i) = n+oo
lim lP'(An I Xo =
i)
= '"'tij
(n)
L
j=1
+ .
r
::::: lim { Ltij(n)
n+oo
} .
Yi
mms:::dYr+s}
j= 1
Yi
''< E.
mins:::dYR+sl
Taker = Rand let n + oo, implying that lP'(A 00
by irreducibility that all states are persistent.
I Xo =
i)
= 0.
j,, nc = 1
= nnttA = nnfnA =
1.
"th
Wl
1 .
= IP'i(TA
< TE).
5
3
1
1
8,
VB = 4> VC = 2, VD = 4
A typical conditional transition probability Tij = IP'i (X 1 = j I TA < TE) is calculated as follows:
SO UtlOn VA
and similarly,
We now compute the conditional expectations Iii = lEi (TA I TA < TE) by conditioning on the first
step of the conditioned process. This yields equations of the form iiA = 1 + ~JiB + ~ Jic, whose
solution gives ji A =
1:.
(e) Either use the stationary distribution of the conditional transition matrix T, or condition on the first
step as follows. With N the number of visits to D, and Tli = lEi (N I TA < TE), we obtain
TIA =
~TIB +~Tic,
whence in particular Tl A =
TIB =
0+
~TIC,
TIC=
0+
TID =TIC,
1~.
285
h, 1TB
= 2.
.t.
[6.5.1][6.5.6]
Solutions
Markov chains
0::::: i::::: b,
an empty product being interpreted as 1. The constant no is chosen in order that the ni sum to 1, and
the chain is therefore timereversible.
2. Letn be the stationary distribution of X, and suppose X is reversible. We have thatniPij = Pjilrj
for all i, j, and furthermore n i > 0 for all i. Hence
3.
With n the stationary distribution of X, look for a stationary distribution v of Y of the form
ififjC,
ifi
C.
(b) i, j
In the limit as f3
+0, the chain Y never leaves the set C once it has arrived in it.
4.
5.
With Yn = Xn  im,
Now iterate.
6. (a) The distribution n1 = f3/(a + /3), n2 = aj(a + /3) satisfies the detailed balance equations,
so this chain is reversible.
(b) By symmetry, the stationary distribution is 1r = (~, ~,~),which satisfies the detailed balance
equations if and only if p =
i.
i.
286
Solutions
[6.5.7][6.6.2]
7. A simple random walk which moves rightwards with probability p has a stationary measure
TCn = A(p/q)n, in the sense that :n: is a vector satisfying :n: = :n:P. It is not necessarily the case that
this :n: has finite sum. It may then be checked that the recipe given in the solution to Exercise (6.5.3)
yields n(i, j) = pfp4 L:(r,s)EC PIp~ as stationary distribution for the given process, where Cis the
relevant region of the plane, and Pi = pi/ qi and Pi (= 1  qi) is the chance that the i th walk moves
rightwards on any given step.
8.
Since the chain is irreducible with a finite state space, we have that TCi > 0 for all i. Assume the
chain is reversible. The balance equations TCi Pij = TCj Pji give Pij = TCj Pji/TCi. Let D be the matrix
with entries 1/ni on the diagonal, and S the matrix (nj Pji ), and check that P = DS.
Conversely, ifP = DS, then di 1 Pij = dj 1 Pji whence TCi = di 1 L:k di: 1 satisfies the detailed
balance equations.
Note that
Pij =
If the chain is reversible in equilibrium, the matrix M = ( y'ni/nj Pij) is symmetric, and therefore M,
and, by the above, P, has real eigenvalues. An example of the failure of the converse is the transition
matrix
P=(i ~
1
!).
0
which has real eigenvalues 1, and! (twice), and stationary distribution :n: = (~,~,b). However,
n1 Pl3 = 0
9.
=f::.
Simply check the detailed balance equations TCi Pij = TCj Pji.
= mr{ "';XiPij}
::::: allxll
where
::::=
Applying the given theorem, there exists a point :n: in C such that T (n) = n, which is to say that
n = nP.
2.
287
if j
::::=
m,
if j = m + 1,
[6.6.3][6.6.4]
Solutions
Markov chains
"'<p
< i _< m,
L
!J  8!J)y
J + ym +I > 0 for 1 _
Ym+l < 0,
j=l
this implies that
m
for all i,
and hence the impossibility that 'L/}=I Pij Yj > maxi {Yi}. It follows that statement (i) holds, which
is to say that there exists a nonnegative vector x = (xl, x2, ... , Xm) such that x(P  I) = 0 and
I:t=l Xi = 1; such an xis the required eigenvector.
3. Thinking of Xn+ 1 as the amount you may be sure of winning, you seek a betting scheme x such
that Xn+ 1 is maximized subject to the inequalities
n
for 1 :S j :Sm.
Xn+l :S LXitij
i=l
Writing aij = tij for 1 :S i :S nand an+l,j = 1, we obtain the linear program:
maximize
Xn+l
subject to
n+l
L Xiaij :S 0
i=l
for 1 :S j :Sm.
minimize
subject to
L aij Yj = 0
j=l
for 1 :S i :S n,
Lan+l,jYj = 1,
j=l
Yj :0::0
for 1 :S j :Sm.
Reexpressing the aij in terms of the tij as above, the dual program takes the form:
m
minimize
subject to
L tij Pj
j=l
=0
for 1 :S i :S n,
L Pj = 1,
j=l
Pj :::: 0
for 1 :S j :Sm.
The vector x = 0 is a feasible solution of the primal program. The dual program has a feasible
solution if and only if statement (a) holds. Therefore, if (a) holds, the dual program has minimal value
0, whence by the duality theorem of linear programming, the maximal value of the primal program is
0, in contradiction of statement (b). On the other hand, if (a) does not hold, the dual has no feasible
solution, and therefore the primal program has no optimal solution. That is, the objective function of
the primal is unbounded, and therefore (b) holds. [This was proved by De Finetti in 1937.]
4. Use induction, the claim being evidently true when n = 1. Suppose it is true for n = m. Certainly
pm+l is of the correct form, and the equation pm+ 1x' = P(Pmx') with x = (1, w, w2 ) yields in its
first row
288
Solutions
[6.6.5][6.7.2]
as required.
5. The first part follows from the fact that Jr1 1 = 1 if and only if 1rU = 1. The second part follows
from the fact that ni > 0 for all i if P is finite and irreducible, since this implies the invertibility of
1P+ U.
6. The chessboard corresponds to a graph with 8 x 8 = 64 vertices, pairs of which are connected
by edges when the corresponding move is legitimate for the piece in question. By Exercises (6.4.6),
(6.5.9), we need only check that the graph is connected, and to calculate the degree of a comer vertex.
(a) For the king there are 4 vertices of degree 3, 24 of degree 5, 36 of degree 8. Hence, the number of
edges is 210 and the degree of a comer is 3. Therefore tt(king) = 420/3 = 140.
(b) tt(queen) = (28 x 21 + 20 x 23 + 12 x 25 + 4 x 27)/21 = 208/3.
(c) We restrict ourselves to the set of 32 vertices accessible from a given comer. Then tt(bishop) =
(14x7+10x9+6x 11+2x 13)/7=40.
(d) tt(knight) = (4 X 2 + 8 X 3 + 20 X 4 + 16 X 6 + 16 X 8)/2 = 168.
(e) tt(rook) = 64 x 14/14 = 64.
7. They are walking on a product space of 8 x 16 vertices. Of these, 6 x 16 have degree 6 x 3 and
16 x 2 have degree 6 x 5. Hence
tt(C) = (6
16
3 + 16
5)/18 = 448/3.
IP All =(A. l)(A. + !)(A.+ ~). Tedious computation yields the eigenvectors, and thus
8.
0 0)
4
1
3
1
1
(n 1) (
1
llD(Z <2yn1Z >0)j[
2.
n
l2ynJ
 1  Gn(O)
"""'
t:J
nk1
(n + l)k+1
=1
1 )  L2ynJ
1+n
+le 2Y.
~ skiP(Zn = k, extinction)
I extmct10n) = L
k=O
IP( extinction)
289
~ skiP(Zn = k)r/
1
= L
= Gn(sry),
k=O
TJ
TJ
[6.7.3][6.8.1]
3.
Markov chains
Solutions
We have that 17 = G(17). In this case G(s) = q(l  ps) 1, and therefore 17 = q I p. Hence
4.
IX
IX
IX
> 0).
(b) Hence
lE(Z~)
Z > 0  lE(Znlttn) 1
)  IP'(Zn > 0) 1 Gn(O)+ 111
I n
n +II N(t)
n)
= JP(A Ll B)
where
We have that
JP(A Ll B) = JP(A) + JP(B)  II"( A
n B)
290
Solutions, [6.8.2][6.8.4]
where C ={two or more flies arrive in (t, t + h]} and D ={two or more wasps arrive in (t, t + h]}.
This probability is no greater than (A.h)(p,h) + o(h) = o(h).
2. Let I be the incoming Poisson process, and let G be the process of arrivals of green insects.
Matters of independence are dealt with as above. Finally,
IP'(G(t +h)= n +II G(t) =
n)
n)
11"( G(t +h) > n +II G(t) = n) ::=:II" (I (t +h) > n +II I (t) = n) = o(h).
3.
1P'(E(t)>xiT1=u)=
IP'(E(tu)>x)
if u :::: t,
ift<u=St+x,
ifu>t+x,
Jo
00
A.eAudu.
t+x
You may solve the integral equation using Laplace transforms. Alternately you may guess the
answer and then check that it works. The answer is IP'(E(t) ::=: x) = I  eh, the exponential
distribution. Actually this answer is obvious since E(t) > x if and only if there is no arrival in
[t, t + x ], an event having probability e Ax.
4.
with boundary conditions Pi} (0) = Oij, the Kronecker delta. We write Gi (s, t) = "E1 si Pi} (t), the
probability generating function of B(t) conditional on B(O) = i. Multiply through the differential
equation by si and sum over j:
a partial differential equation with boundary condition Gi (s, 0) = si. This may be solved in the usual
way to obtain Gi(s, t) = g(eAt(l s 1)) for some function g. Using the boundary condition, we
find that g(I s 1) = si and so g(u) = (1  u)i, yielding
I
(seAt)i
Gi(s, t) = {I eAI(I s1 )}i = {I s(l eAt)}i.
as required.
291
?:_
i,
[6.8.5][ 6.8.6]
Solutions
Markov chains
Alternatively use induction. Set j =ito obtain pji(t) = Aipu(t) (remember Pi,i1 (t) = 0),
and therefore Pi i (t) = e Ai t. Rewrite the differential equation as
d ( Pij(t)e A1't) = A(j l)Pi,jt(t)e A1't .
dt
Set j = i + 1 and solve to obtain Pi,i+l (t) = ieAit ( 1 eAt). Hence(*) holds, by induction.
The mean is
E(B(t)) = !_GJ(S,
as
t)l
=/eM,
s=l
as
s=l
Alternatively, note that B(t) has the negative binomial distribution with parameters eAt and I.
5.
n 2: 0,
where Ai = i A+ v. The process is honest, and therefore m(t) = L::n npn(t) satisfies
00
00
+ v]Pnt(f)
n=l
Ln(nA + v)pn(t)
n=O
00
= L
n=O
00
L(An
n=O
+ v)pn(t)
= Am(t)
+ v.
6. Using the fact that the time to the nth arrival is the sum of exponential interarrival times (or using
equation (6.8.15)), we have that
is given by
where
n
az
A.
IT1AA
j=O 1
Hi
292
Solutions
[6.8.7][6.9.1]
so long as Ai =!= A.J whenever i =!= j. The Laplace transform Pn may now be inverted as
limn~oo
.
=ITn __). z= lE(eeTn)
+ ()
Xo +X 1 + + Xn where Xk is the (k + l)th interarrival time, a random variable which is
i=O Ai
since Tn =
exponentially distributed with parameter Ak Using the continuity theorem, lE(eeTn) + 1E(eeT) as
n+ oo, whence AnPn(()) + 1E(eeT) as n+ oo, which may be inverted to obtain AnPn(t) + f(t)
as n+ oo where f is the density function ofT. Now
+ ~) 2 , we have that
lE(eeT) =
IT
00
n=O
1+
() 1 2
(n + 2)
}1
= sech
(nv'e).
Inverting the Laplace transform (or consulting a table of such transforms) we find that
where P12 = 1  Plio P21 = 1  P22 Solve these subject to Pi} (t) = 8ij, the Kronecker delta, to
obtain that the matrix P1 = (Pi} (t)) is given by
(b) There are many ways of calculating Gn; let us use generating functions. Note first that G0 =I,
the identity matrix. Write
n ::=: 0,
293
[6.9.2][6.9.4]
Solutions
Markov chnins
Hence an+1 = (pfA.)cn+1 for n 2: 0, and the first difference equation becomes an+1 = (A.+ p,)an,
n 2: 1, which, subject to a1 = p,, has solution an = (l)np,(A. + p,)n 1, n 2: 1. Therefore
Cn = (l)n+ 1A.(A. + p,)n 1 for n 2: 1, and one may see similarly that bn =an, dn = en for
n 2: 1. Using the facts that ao =do = 1 and bo = co = 0, we deduce that L:~0 (tn jn!)Gn = Pt
where P 1 is given in part (a).
(c) With1r = (n1, n2), we have that p,n1 + A.n2 = 0 and p,n1 A.n2 = 0, whence n1 = (A.jp,)n2.
In addition, n1 + n2 = 1 if n1 = A./(A. + p,) = 1 n2.
2.
1)
p!2(t)P21 (2t)
Pll (3t)
Pu (3t)pu (t)
3. The interarrival times and runtimes are independent and exponentially distributed. It is the lackofmemory property which guarantees that X has the Markov property.
The state space isS= {0, 1, 2, ... } and the generator is
G~
c
~
A.
(A.+ p,)
p,
0
A.
(A.+ p,)
0
0
A.
...
with solutionni = no(A.jp,)i. We have in addition that:Li ni = 1 ifA < p,andno = {1 (A./p,)} 1.
4.
One may use the strong Markov property. Alternatively, by the Markov property,
lP'(Yn+1 = j I Yn = i, Tn = t, B)= lP'(Yn+1 = j I Yn = i, Tn = t)
I Yn =
I Yn =
i, Tn = t)fTn(t)dt
= lP'(Yn+1 = j I Yn = i),
so that Y is a Markov chain. Now qij = lP'(Yn+1 = j I Yn = i) is given by
% = { 00 Pij(t)A.eM dt,
lo
294
Solutions
[6.9.5][6.9.8]
by conditioning on the (n + 1)th interarrival time of N; here, as usual, Pij (t) is a transition probability
of X. Now
The jump chain Z = {Zn : n ::=: 0} has transition probabilities hij = 8i} / 8i, i =f. j. The chance
that Z ever reaches A from j is also IJj, and 11} = Lk hjkiJk for j </:. A, by Exercise (6.3.6). Hence
gjiJj = Lk 8jk11k> as required.
5.
6. Let T1 = inf {t : X (t) =f. X (0)}, and more generally let Tm be the time of the mth change in value
of X. For j :. A,
/Jj = 1Ej(T1)
+ Lhjk/Jk>
kfj
where Ej denotes expectation conditional on Xo = j. Now Ej (Tl) = g1 1 , and the given equations
follow. Suppose next that (ak : k E S) is another nonnegative solution of these equations. With
Ui = Ti+1  Ti and R = min{n 2: 1 : Zn E A}, we have for j </:.A that
where
2: Ej(Uo)
= Ej (
+ 1Ej(U1/{R>l)) + + Ej(Unl(R>nJ)
Url(R>r})
= Ej (min{Tn, HA})
+ Ej(HA)
r=O
7. First note that i is persistent if and only if it is also a persistent state in the jump chain Z. The
integrand being positive, we can write
where {Tn : n ::=: 1} are the times of the jumps of X. The right side equals
00
00
gz n=O
where H = (hij) is the transition matrix of Z. The sum diverges if and only if i is persistent for Z.
8. Since the imbedded jump walk is persistent, so is X. The probability of visiting m during an
excursion is a= (2m) 1, since such a visit requires an initial step to the right, followed by a visit to
m before 0, cf. Example (3.9.6). Having arrived at m, the chance of returning tom before visiting 0
is 1 a, by the same argument with 0 and m interchanged. In this way one sees that the number N
of visits tom during an excursion from 0 has distribution given by IP'(N ::=: k) = a(l  a)k 1, k ::=: 1.
The 'total jump rate' from any state is A., whence T may be expressed as :L~o Vi where the Vi are
exponential with parameter A.. Therefore,
JE(e eT )=GN ( A. )
).()
a).
=(1a)+a.
a).()
295
[6.9.9][6.9.11]
Solutions
Markov chains
The distribution ofT is a mixture of an atom at 0 and the exponential distribution with parameter a)..
9. The number N of sojourns in i has a geometric distribution IP'(N = k) = fkl(l f), k =:: 1,
for some f < 1. The length of each of these sojourns has the exponential distribution with some
parameter 8i. By the independence of these lengths, the total time T in state i has moment generating
function
gi(lf)
8i0 f ) 
10. The jump chain is the simple random walk with probabilities ).j (). + JJ) and JJ/ (). + JJ), and with
POl = 1. By Corollary (5.3.6), the chance of ever hitting 0 having started at 1 is JJ/A, whence the
probability of returning to 0 having started there is f = JJ/A. By the result of Exercise (6.9.9),
as required. Having started at 0, the walk visits the state r =:: 1 with probability 1. The probability of
returning to r having started there is
and each sojourn is exponentially distributed with parameter 8r = ). + JJ. Now g 7 (1  fr) = ).  JJ,
whence, as above,
JE(eevr) =
).  /JAJJe
The probability of ever reaching 0 from X (0) is (JJ/A)x (O), and the time spent there subsequently
is exponential with parameter ).  JJ. Therefore, the mean total time spent at 0 is
where we have used the fact thatJrG = 0. Also irk =:: 0 and L:k irk = 1, and thereforeJi is a stationary
distribution of Y.
Clearly fik = nk for all k if and only if 8k = l::i ni 8i for all k, which is to say that 8i = 8k for
all pairs i, k. This requires that the 'holding times' have the same distribution.
(b) Let Tn be the time of the nth change of value of X, with To= 0, and let Un = Tn+l  Tn. Fix a
state k, and let H = min{n =:: 1 : Zn = k}. Let Yi (k) be the mean time spent in state i between two
consecutive visits to k, and let Yi(k) be the mean number of visits to i by the jump chain in between two
296
Solutions
[6.9.12][6.11.2]
visits to k (so that, in particular, Yk(k) = g/; 1 and n(k) = 1). With Ej and 1P'j denoting expectation
and probability conditional on X (0) = j, we have that
Yi(k) =lEk(f:unl(Zn=i,H>n}) = f:Ek(Un I f(Zn=iJ)lP'k(Zn = i, H > n)
n=O
n=O
1
1
lP'k(Zn = i, H > n) = yi(k).
n=O 8i
8i
00
= L
The vector y(k) = (Yi (k) : i E S) satisfies y(k)H = y(k), by Lemma (6.4.5), where H is the
transition matrix of the jump chain Z. That is to say,
for j E S,
whence I:i Yi (k)gij = 0 for all j E S. If tlk = I:i Yi (k) < oo, the vector (Yi (k)/ ttk) is a stationary
distribution for X, whence :rri = Yi(k)/ttk for all i. Setting i = k we deduce that Trk = 1/(gktLk).
Finally, ifi;i Tri8i < oo, then
tlk =::::: =
Trk
I:i Tri8i
"'
= tlk L:rri8i
Trk8k
i
12. Define the generator G by gu = vi, 8ij = vihij, so that the imbedded chain has transition
matrix H. A root of the equation 1rG = 0 satisfies
0 = L:rri8ij = TrjVj
i
L (:rrivi)hij
i:i/j
whence the vectors = (:rrjVj : j E S) satisfies s = tH. Therefore s = av, which is to say that
Trj Vj = avj ,for some constant a. Now Vj > 0 for all j, so that Trj = a, which implies that I:j Trj =!= 1.
Therefore the continuoustime chain X with generator G has no stationary distribution.
The jump chain is a walk {Zn} on the setS= {0, 1, 2, ... } satisfying, fori ::: 1,
lP'(Zn+I = j
I Zn = i) = {
if j = i + 1,
ifj=i1,
Pi
1 Pi
'f .
1 ]=!'
ifj=i+l.
To find the stationary distribution of Y, either solve the equation 1r = 1rQ, or look for a solution of
the detailed balance equations :rrihi,i+l = :rri+Ihi+ 1,i. Following the latter route, we have that
297
[6.11.3][6.11.4]
Solutions
Markov chains
whence TCi =reo pi (1 + i j p )/ i! fori 2:: 1. Choosing reo accordingly, we obtain the result.
It is a standard calculation that X has stationary distribution v given by vi = pie p j i! fori 2:: 0.
The difference between 1r and v arises from the fact that the holdingtimes of X have distributions
which depend on the current state.
3.
subject to 11 (0) = 0.
Rewrite the equation as
11'
,,,___:_, = 1
(1  IJ)(JJ  A.IJ)
4.
lP'(X(t) = 0)
lP'(X(u) = 0)
17(t)
= .
17(u)
as usual. The generating function of X(t), conditional on {X(t) > 0}, is therefore
lP'(X(t) > 0)
H(s)
1  G(O, t)
oo to obtain as limit
(M A.)s
/)AS
oo
= Ls
Pn
n=l
298
Solutions
Special processes
5.
[6.11.5][6.12.1]
IT
< oo)
!oo
oo {
)..
1  lE(sX(t))l =O
M
s
dt
1
( ).. )
log   .
M
A.  M
In the case A.= JJ, lP'(T < oo) = 1 and JE(T) = oo.
6. By considering the imbedded random walk, we find that the probability of ever returning to 1
is max{A., JJ}/(A. + JJ), so that the number of visits is geometric with parameter min{A., JJ}/(A. + JJ).
Each visit has an exponentially distributed duration with parameter A. + JJ, and a short calculation
using moment generating functions shows that V1 (oo) is exponential with parameter min{A., JJ}.
Next, by a change of variables, Theorem (6.11.10), and some calculation,
~srlE(Vr(t)) = lE(~
sr f(X(u)=r)du) = lE
(l
sX(u) du)
1
{ A.(l s) (M A.s)e(A~L)t}
loot lE(s X( ) du = JJfA.  log
A.
A.JJ1 { A.s(eP 1)} + terms not involving s,
= log 1 
u )
1 
A.
JJeP A.
where p = JJ  A.. We take the limit as t + oo and we pick out the coefficient of sr.
7.
A.t(1  s)
A.t(l  s)
+s = 1 _
+1
1 s
A.t(1  s) + 1
and
r lE(sX(u)) du
lo
= t  _!_ log{A.t(l s)
A.
+ 1}
1 1og { 1  Ats
= }
A.
1 +A.t
.
. s.
+ terms not mvolvmg
Letting t+ oo and picking out the coefficient of sr gives JE(Vr(oo)) = (rA.) 1 . An alternative
method utilizes the imbedded simple random walk and the exponentiality of the sojourn times.
299
The
[6.12.2][6.12.4]
Solutions
Markov chains
whence JJ,lQ = (JJ, + A.)/(JJ, A.). Since each sojourn is exponentially distributed with parameter
JJ, +A., the result follows by an easy calculation. See also Theorem (11.3.17).
2.
the probability generating function of the population size at time u in a simple birth process. In the
absence of disasters, the probability generating function of the ensuing population size at time v is
The individuals alive at timet arose subsequent to the most recent disaster at timet  D, where D
has density function 8e 8x, x > 0. Therefore,
3. The mean number of descendants after time t of a single progenitor at time 0 is e<A~L)t. The
expected number due to the arrival of a single individual at a uniformly distributed time in the interval
on [0, x] is therefore
1
e(A!L)X  1
e(A~L)u du = ,,x 0
(J... JJ,)X
lox
The aggregate effect at timex of N earlier arrivals is the same, by Theorem (6.12.7), as that of N
arrivals at independent times which are uniformly distributed on [0, x]. Since JE(N) = vx, the mean
population size at timex is v[e(A~L)x 1]/(A. JJ,). The most recent disaster occurred at timet D,
where D has density function 8e  8x, x > 0, and it follows that
lE(X(t)) =
loo
8e 8x[e(A11)x
A.JJ,
1]dx
v
8 x[e(A~L)t 1].
+ e
A.JJ,
4.
Let N be the number of clients who arrive during the interval [0, t]. Conditional on the event
{N = n}, the arrival times have, by Theorem (6.12.7), the same joint distribution as n independent
variables chosen uniformly from [0, t]. The probability that an arrival at a uniform time in [0, t] is
still in service at timet is f3 = f~[l  G(t x)]t 1 dx, whence, conditional on {N = n}, the total
number M still in service is bin(n, {J). Therefore,
whence M has the Poisson distribution with parameter A.{Jt = A. f~[l  G(x)] dx. Note that this
parameter approaches A.JE(S) as t + oo.
300
Solutions
[6.13.1][6.13.3]
lo
f3 + Y
(b) Using (a), and the lackofmemory of the process, the required probability is
lE(min{S, T} B(l)
1
}
G(1)+2
= 1) = lE {
.
y 1 + eY
y2
2. Let B, be the ball with centre 0 and radius r, and let Nr = ITI
(6.13.11) that S, = Z::xEnnB, g(x) satisfies
JE(S, IN,)= N,
where A(B) = fyEB A.(y) dy. Therefore, JE(S,)
gence that JE(S) = JJRd g(u)A.(u) du. Similarly,
JE(s;
N,) = lE
(l L
1
Br
n Brl
We have by Theorem
A.(u)
g(u)du,
A(B,)
g(x)J )
EnnBr
~ c~H,
E
g(x)
2))
+E (
g(x)g(y))
x,yEnnBr
= N,
A.(u)
g(u)   du
Br
whence
lE(S;) =
A(Br)
+ N, (Nr
 1)
11u
g(u)g(v)
u,vEBr
A.(u)A.(v)
A(Br)
2 du dv,
By monotone convergence,
301
[6.13.4][6.13.8]
Solutions
Markov chains
elementary coordinate geometry, the intensity function in plane polar coordinates is 2'A/ ~.
< 2n.
The Mercator projection represents the spherical coordinates (e, )as Cartesian coordinates in
the range 0 ::::; < 2n, 0 ::::;
n . (Recall that is the angle made with the axis through the north
pole.) Therefore a uniform intensity on the globe corresponds to an intensity .function 'A sine on the
map. Likewise, a uniform intensity on the map corresponds to an intensity 'A/ sine on the globe.
5.
e ::;
6.
Let the Xr have characteristic function. Conditional on the value of N(t), the corresponding
arrival times have the same distribution as N(t) independent variables with the uniform distribution,
whence
JE(eieS(t)) = lE{lE(eieS(t) I N(t))} = lE{lE(eiexeau )N(t)}
lot {(ee"u) 1} du }
Now, for s < t, S(t) = S(s )ea(ts) + S(t s) where S(t s) is independent of S(s) with the same
distribution as S(t s). Hence, for s < t,
'AlE(X 2 )
'AJE(X 2 )
cov(S(s), S(t)) = ~a(ts) = ~(1 e2as)ea(ts)+ ~eav
ass + oo with v = t  s fixed. Therefore, p(S(s), S(s + v))+ eav ass+ oo.
7.
Differentiate with respect to x andy to obtain the joint density function A.(x)'A(x
+ y)eA(x+y),
x, y ::: 0. Since this does not generally factorize as the product of a function of x and a function of y,
T1 and Tz are dependent in general.
8.
Let Xi be the time of the first arrival in the process Ni. Then
lP'(I
=I, T:::
t)
= lP'(t::::; x 1
00
< inf{X 2 ,
x3 , ... })
302
Al eJ..t_
'A
Solutions
[6.14.1][6.14.4]
If P is reversible then
RHS
l,]
l,]
~njXj (~PjiYi)
J
= LHS.
Suppose conversely that (x, Py) = (Px, y) for all x, y E l 2 (n). Choose x, y to be unit vectors with 1
in the ith and jth place respectively, to obtain the detailed balance equations ni Pij = nj Pji.
2. Just check that 0 ::::; bij ::::; 1 and that the Pij = 8ij bij satisfy the detailed balance equations
(6.14.3).
3. It is immediate that Pjk = IAjkl, the Lebesgue measure of Ajk This is a method for simulating
a Markov chain with a given transition matrix.
4. (a) Note first from equation (4.12.7) that d(U) = ~ supi#j dTv(uio Uj.), where Ui. is the mass
function Uif, t E T. The required inequality may be hacked out, but instead we will use the maximal
coupling of Exercises (4.12.4, 5); see also Problem (7.11.16). Thus requires a little notation. For
i, j E S, i =f. j, we find a pair (Xi, Xj) of random variables taking values in T according to the
marginal mass functions Ui., Uj., and such that IP'(Xi =f. Xj) = ~dTv(uh Uj.). The existence of
such a pair was proved in Exercise (4.12.5). Note that the value of Xi depends on j, but this fact
has been suppressed from the notation for ease of reading. Having found (Xi, Xj ), we find a pair
(Y(Xi), Y(Xj)) taking values in U according to the marginal mass functions vx; vxj'' and such that
IP'(Y(Xi)
=f.
Y(Xj)
I Xi,
Xj) =
IP'(Y(Xi)
=f.
Y(Xj)~=
IP'(Xi = r, Xj = s)IP'(Y(r)
=f. Y(s))
,sES
is
= L
IP'(Xi = r, Xj = s)idTV(Vr., Vs.)
r,sES
ri=s
::'0
whence
d(UV) = supiP'(Y(Xi)
=f.
Y(Xj)) ::'0
ii=j
g rfs
supdTV(Vr., Vs.) }{sup!P'(Xi =f. Xj)}
i,j
u=
(IP'(Xo = 1)
IP'(Yo = 1)
IP'(Xo = 2)
IP'(Yo = 2)
m))
IP'(Xo =
IP'(Yo = m)
303
[6.15.1][6.15.6]
Solutions
Marlwv chains
2.
whence, by induction, the nstep transition matrix pn is doubly stochastic for all n :::: 1.
If j is not nonnull persistent, then Pij (n) + 0 as n + oo, for all i, implying that l::i Pij (n) + 0,
a contradiction. Therefore all states are nonnull persistent.
If in addition the chain is irreducible and aperiodic then Pij (n) + lij, where 1f is the unique
stationary distribution. However, it is easy to check that 1f = (N 1 , N 1 , ... , N 1) is a stationary
distribution if Pis doubly stochastic.
(b) Suppose the chain is persistent. In this case there exists a positive root of the equation x = xP, this
root being unique up to a multiplicative constant (see Theorem (6.4.6) and the forthcoming Problem
vector of 1's. By the
(7)). Since the transition matrix is doubly stochastic, we may take x = 1, ~e
above uniqueness of x, there can exist no stationary distribution, and there e the chain is null. We
deduce that the chain cannot be nonnull persistent.
3.
Choose two states i and j, and pick m and n such that a= Pij (m)Pji(n) > 0. Then
Pu(m
+ r + n):::: apjj(r).
Set r = 0 to find that Pii (m + n) > 0, and so d(i) I (m + n). If d(i) f r then Pii (m + r
that Pjj(r) = 0; therefore d(i) I d(j). Similarly d(j) I d(i), giving that d(i) = d(j).
+ n) =
0, so
n)
= Pir(n)Pjs(n) > 0
if n:::: max{N(i, r), N(j, s)}, so that the chain is irreducible and aperiodic.
(c) SupposeS= {1, 2} and
P=C 6)
In this case { {1, 1}, {2, 2}} and { {1, 2}, {2, 1}} are closed sets of states for the bivariate chain.
5. Clearly lP'(N = 0) = 1  /ij, while, by conditioning on the time of the nth visit to j, we
have that lP'(N :::: n + 1 I N :::: n) = fjj for n :::: 1, whence the answer is immediate. Now
lP'(N = oo) = 1  l:~o lP'(N = n) which equals 1 if and only if /ij = fjj = 1.
6. Fix i =/= j and let m = min{n : Pij (n) > 0}. If Xo = i and Xm = j then there can be no
intermediate visit to i (with probability one), since such a visit would contradict the minimality of m.
304
Solutions
Problems
[6.15.7][6.15.8]
Suppose Xo = i, and note that (1  f}i) Pij (m) :5 1  Iii, since if the chain visits j at time m
and subsequently does not return to i, then no return to i takes place at all. However fii = 1 if i is
persistent, so that fj i = 1.
7.
(a) We may takeS= {0, 1, 2, ... }. Note that qij (n)::: 0, and
00
L%(n) =
%(n
1,
+ 1) = Lqu(l)qu(n),
1=0
whence Q = (% (1)) is the transition matrix of a Markov chain, and Qn = (% (n)). This chain is
persistent since
for all i,
n
and irreducible since i communicates with j in the new chain whenever j communicates with i in the
original chain.
That
i =I= j, n ::: 1,
is evident when n = 1 since both sides are qij (1). Suppose it is true for n = m where m ::: 1. Now
lji(m
+ 1) =
i =I= j,
ljk(m)Pki
k:kfj
so that
_l_ lji(m
+ 1) =
Xi
i =I= j,
gkj(m)qik(1),
k:kfj
where Pi(j) is the mean number of visits to i between two visits to j; we have used the fact that
I:n gij (n) = 1, since the chain is persistent (see Problem (6.15.6)). It follows that Xi = XOPi (0) for
all i, and therefore x is unique up to a multiplicative constant.
(c) The claim is trivial when i = j, and we assume therefore that i =I= j. Let Ni (j) be the number
of visits to i before reaching j for the first time, and write lP'k and lEk for probability and expectation
conditional on Xo = k. Clearly, lP'j(Ni(j):::: r) = hjiO hijyl for r::: 1, whence
8.
Un =
fiUni
i=l
305
n:::
1,
[6.15.9][6.15.9]
where
Solutions
Marlwv chains
fi is the probability that the first return of X to its persistent starting point s takes place at time
i. Certainly uo = 1.
Conversely, suppose u is a renewal sequence with respect to the collection Um : m 2: 1). Let X
be a Markov chain on the state space S = {0, 1, 2, ... } with transition matrix
.. _ { lP'(T 2: i + 2 I T 2: i + 1)
PI] 1  lP'(T 2: i + 2 I T 2: i + 1)
if j = i + 1,
if j = 0,
where Tis a random variable having mass function fm = lP'(T = m). With Xo = 0, the chance that
the first return to 0 takes place at time n is
lP'( Xn = 0,
IT
xi
i= 0 I Xo
IT
= ( 1 _ G(n + 1))
G(i + 1)
G(n)
i=l
G(i)
= G(n) 
G(n
+ 1) =
fn
vo
= 1,
forn::;::1,
Vn = LfiVni
i=l
2n
(2n)
9. Of the first
steps, let there be i rightwards, j upwards, and k inwards. Now Xzn = 0 if
and only if there are also i leftwards, j downwards, and k outwards. The number of such possible
!/ {(i! j! k!) 2 }, and each such combination has probability (t) 2U+ j+k) = (t) 2
combinations is
The first equality follows, and the second is immediate.
Now
n.
1 ) 2n
lP'(Xzn = 0) ::S ( 2
where
M=max{ 3 n~!lkl:
l. J . .
(2n) M
n
i+J+k=n
n!
3n1 "lkl
l.J.
i,j,k::;::O, i+j+k=n}.
It is not difficult to see that the maximum M is attained when i, j, and k are all closest to :} n, so that
Furthermore the summation in (*) equals 1, since the summand is the probability that, in allocating n
balls randomly to three urns, the urns contain respectively i, j, and k balls. It follows tliat
(2n)!
lP'(X2 = 0) < '7::n
 12nn!
306
(LjnJ !)3
Solutions [6.15.10][6.15.13]
Problems
which, by an application of Stirling's formula, is no bigger than Cn 2 for some constant C. Hence
l:n lP'(X2n = 0) < oo, so that the origin is transient.
10. No. The line of ancestors of any cellstate is a random walk in three dimensions. The difference
between two such lines of ancestors is also a type of random walk, which in three dimensions is
transient.
11. There are one or two absorbing states according as whether one or both of a and f3 equal zero. If
af3 =f. 0, the chain is irreducible and persistent. It is periodic if and only if a = f3 = 1, in which case
it has period 2.
(a!f3'a:f3)
is the stationary distribution. There are various ways of calculating pn; see Exercise (6.3.3) for
example. In this case the answer is given by
Pi}=
(NNi)2
ifj=i+1,
(NN i) 2 if j = i,
1  (Ni )
if j = i  1,
(Ni )2
2
for 0 :::; i :::; N. This process is a birthdeath process in discrete time, and by Exercise (6.5.1) is
reversible in equilibrium. Its stationary distribution satisfies the detailed balance equation Jri Pi,i+l =
ni+lPi+l,i for 0:::; i < N, whence ni = no(~) 2 for 0:::; i :::; N, where
_!_
no
t (~)
i=O
(2N).
N
13. (a) The chain X is irreducible; all states are therefore of the same type. The state 0 is aperiodic,
and so therefore is every other state. Suppose that Xo = 0, and let T be the time of the first return to
0. Then lP'(T > n) = aoa1 an1 = bn for n :::: 1, so that 0 is persistent if and only if bn + 0 as
n + oo.
(b) The mean of T is
00
JE(T)
00
[6.15.14][6.15.14]
Solutions
Markov chains
n1
II (1 
bn =hi
Aif3),
n 2: /.
i=I
Hence bn + 0 if and only if :Li AifJ = oo, which is to say that
persistent if and only if f3 ::=: 1.
(d) We have that 1  x :::: ex for x 2: 0, and therefore
f3 ::=:
if{J<l.
bn = b I
n1(
A)
II
1  ;
i=I
2: b I
n1(.
II ~1)
i=I
= bI
(/ =1) .
n
IPi} (t + h) 
Pij (t) I =
IL (Pik (h) 
I :::: (1 
::S
(1 
+ L Pik (h)
k=Ji
k
Pii (h))
+ (1 
1
((Pii(t)  1)
g(t)
Pii (t)  1
t log{1 (1 pu(t))}+ A.
308
Solutions
Problems
[6.15.15][6.15.16]
15. Let i and j be distinct states, and suppose that Pij (t) > 0 for some t. Now
Pij (t) = t
n"'l
L..J k! t (G )ij
k=n
is strictly positive for all sufficiently small positive values oft. Therefore i communicates with j.
16. (a) Suppose X is reversible, and let i and j be distinct states. Now
lP'(X(O) = i, X(t) =
j)
= lP'(X(t) = i, X(O) =
j),
which is to say that Tri Pij (t) = Trj Pji (t). Divide by t and take the limit as t
+0 to obtain that
Trigij = Trjgji
Suppose now that the chain is uniform, and X (0) has distribution 1r. If t > 0, then
lP'(X (t) = j) =
so that X(t) has distribution 1r also. Now lett < 0, and suppose that X(t) has distribution J.L. The
distribution of X (s) for s ::: 0 is J.LPst = 1r, a polynomial identity in the variables  t, valid for all
s ::: 0. Such an identity must be valid for all s, and particularly for s = t, implying that J.L = 1r.
Suppose in addition that Tri gij = Trj gji for all i, j. For any sequence k1, k2, ... , kn of states,
Trigi,ktgkt,kz gkn,j = gkJ,iTrktgkt,kz gkn,j = = gk1,igkz,k1 gj,knTrj.
Sum this expression over all sequences k1, k2, ... , kn oflength n, to obtain
Tri(Gn+1)ij = Trj(Gn+1)ji
n::: 0.
309
[6.15.17][6.15.19]
Markov chains
Solutions
=(a a)
f3
{3
f3 ) .
y
+ f3h(t)
y(1 h(t))
)P = ( y
y
{3(1 h(t)))
f3 + yh(t)
= a(f3 +
y), which is
aa
at
=(s 1)(A.G 11
no::
I.
aa)
as
ds
dt=   JL(S  1)
dG
A.(s  1)G'
and therefore G = eP(si) f ( (sl)eJU), for some function f, determined by the boundary condition
to satisfy eP(sl) f(s 1) =sf. The claim follows.
As t+ oo, G(s, t)+ eP(si), the generating function of the Poisson distribution, parameter p.
310
i < j.
Solutions [6.15.20][6.15.20]
Problems
Assume N(s) = i and s < t. In the usual way,
00
G(s, t; x)
= 2:xilP'(N(t) = j
I N(s)
= i)
J=i
satisfies
iJG
1t
A.(u) du},
whence Pi} (t) is found to be the probability that A = j  i where A has the Poisson distribution with
J:
using the fact that Pi+ I,} (t) = Pi,J1 (t), we are led to
iJG
  = A.(s)(x !)G.
OS
= Poo(t) = exp {
so that
fT(t) = A.(t)exp {
l
lot
A.(u) du},
t 2:0.
A.(u)du }
+ t),
E(T)
= !o oo lP'(T
0
> t)dt
= !ooo
o
du
(1
+ u)c
1 F(u)
,
I  F(s)
0<
S ::": U,
311
0:::: u:::: s.
[6.15.21][6.15.21]
Solutions
Markov chains
IP'(XM > u, B
IM =
m)IP'(M = m)
m=i
00
IM =
IP'(XM > u
m)IP'(B
IM =
m)IP'(M = m)
m=i
00
= IP'(XM > u)
L IP'(B I M = m)IP'(M = m)
m=i
0<
S _:::: U,
where we have used the fact that IP'(XM > u I M = m) is independent of m. It follows that the first
record value exceeding s is independent of all record values not exceeding s. By a similar argument
(or an iteration of the above) all record values exceeding s are independent of all record values not
exceeding s.
The chance of a record value in (s, s + h] is
IP'(s < XM < s +h)=

1 F(s)
J(s)h
1 F(s)
+o(h).
A very similar argument works for the runnersup. Let XM 1 , XM2 , .. be the values, in order,
of offers exceeding s. It may be seen that this sequence is independent of the sequence of offers
not exceeding s, whence it follows that the sequence of runnersup is a nonhomogeneous Poisson
process. There is a runnerup in (s, s + h] if (neglecting terms of order o(h)) the first offer exceeding
s is larger than s + h, and the second is in (s, s + h]. The probability of this is
( 1F(s+h)) (F(s+h)F(s)) +o(h ) = f(s)h +o (h) .
1F(s)
1F(s)
1F(s)
21. Let F 1 (x) = IP'(N*(t) _:::: x), and let A be the event that N has a arrival during (t, t +h). Then
Fr+h(x) = A.h!P'(N*(t +h)_::::
where
IP'(N*(t +h)_::::
Hence
~Fr(x) =
at
xI A)=
A.Fr(x)
L:
t.../
00
oo
Fr(x y)f(y)dy.
Fr(x y)f(y)dy.
312
Solutions [6.15.22][6.15.25]
Problems
I A)}=
1{eJ.It(s1)
+ eJ.2t(sl)},
whence E(N(t))
23.
n 2: 0,
with the convention that Pi (t) = 0. Multiply by sn and sum to deduce that
(1
aa
aa
aG
2
+ J.Lf)at = sG + JLS as  G JLSas
as required.
Differentiate with respect to s and take the limit ass
m(t)
= E(X(t)) = aal
as
s=i
satisfies (1 + J.Lt)m 1(t) = 1 + J.Lm(t) subject to m(O) =I. Solving this in the usual way, we obtain
m(t) = I+ (1 + JL/)t.
Differentiate again to find that
n(t) = E(X(t)(X(t)
satisfies (1
+ J.Lf)n 1 (t)
= 2(m(t)
+ J.Lm(t) + J.Ln(t))
n(t) = I (I  1)
1))
a2al
= 
as
s=i
+ m(t) 
m(t) 2 .
AJ
"}
as required. Set Xi
= TJi+i
ILJ
"}
XJ
!Lj
= JLjXJi
for j 2: 1, so that
=xoiT
f.
i=i
l
It follows that
TJ}+i
= TJO +
k=O
k=O
L Xk = 1 + (TJJ  1) L ek.
The TJ} are probabilities, and lie in [0, 1]. If l::f ek = oo then we must have TJJ = 1, which implies
that TJ} = 1 for all j.
313
[6.15.26][6.15.28]
Solutions
Markov chains
(b) By conditioning on the first step, the probability T/J, of visiting 0 having started from j, satisfies
u + I) 2 TJj+i + p,j1
j2
T/j =
+ (j + 1)2
Hence, (j + 1) 2 (TJj+i TJj) = P(TJj T/jJ), giving (j + 1) 2 (TJj+i TJj) = TJi TJO Therefore,
j
k=O (k
+ 1)
I 2
2 + (1 TJI)6n
as j+ oo.
By Exercise (6.3.6), we seek the minimal nonnegative solution, which is achieved when T/1 = 1 (6/n 2).
26. We may suppose that X(O) = 0. Let Tn = inf{t : X(t) = n}. Suppose Tn = T, and let
Y = Tn+i  T. Condition on all possible occurrences during the interval (T, T +h) to find that
E(Y) = (A.nh)h + f.lnh(h + E(Y 1)) + (1  A.nh P,nh)(h + E(Y)) + o(h),
where Y' is the mean time which elapses before reaching n + 1 from n 1. Set mn = E(Tn+i  Tn)
to obtain that
mn = P,nh(mni + mn) + mn + h{l (A.n + P,n)mn} + o(h).
Divide by h and take the limit ash
.,!..
1
f.ln
1
Jln
Jlnf.lni Ill
mn =  + mni = =  +    + +
An
An
An
An Ani
An Ani A.o
since mo = A. 01 . The process is dishonest if ~~ 0 mn < oo, since in this case T00 = lim Tn has
finite mean, so that lP'(T00 < oo) = 1.
On the other hand, the process grows no faster than a birth process with birth rates Ai, which is
honest if ~~ 0 1/A.n = oo. Can you find a better condition?
27. We know that, conditional on X (0) = I, X (t) has generating function
A.t(ls)+s)I
G(s, t) = ( A.t(l  s) + 1
,
so that
lP'(T:::; x
A.x
X(O) =I)= lP'(X(x) = 0 I X(O) =I)= G(O, x) = ( A.x + 1
)I
)I
lP'(X(O) = /) = Gx(O)
A.x ) + 1.
Ax + 1
For the final part, the required probability is {xI I (xI + 1)} I = { 1 + (x /)I }I, which tends to
eijx as I + oo.
28. Let Y be an immigrationdeath process without disasters, with Y (0) = 0. We have from Problem
(6.15.18) that Y(t) has generating function G(s, t)
exp{p(s 1)(1 e!Lt)} where p A.fp,. As
seen earlier, and as easily verified by taking the limit as t + oo, Y has a stationary distribution.
314
Solutions [6.15.29][6.15.31]
Problems
From the process Y we may generate the process X in the following way. At the epoch of each
disaster, we paint every member of the population grey. At any given time, the unpainted individuals
constitute X, and the aggregate population constitutes Y. When constructed in this way, it is the case
that Y(t) :::; X(t), so that Y is a Markov chain which is dominated by a chain having a stationary
distribution. It follows that X has a stationary distribution :n: (the state 0 is persistent for X, and
therefore persistent for Y also).
Suppose X is in equilibrium. The times of disasters form a Poisson process with intensity 8. At
any given time t, the elapsed time T since the last disaster is exponentially distributed with parameter
8. At the time of this disaster, the value of X (t) is reduced to 0 whatever its previous value.
It follows by averaging over the value ofT that the generating function H(s) = L~o snnn of
X (t) is given by
29. Let G(IBI, s) be the generating function of X(B). If BnC = 0, then X(B UC) = X(B)+X(C),
so that G(a + f.l, s) = G(a, s)G(f.l, s) for Is I :::; 1, a, f.l ::=: 0. The only solutions to this equation which
are monotone in a are of the form G(a, s) = eaJ..(s) for lsi :::; 1, and for some function A.(s). Now any
interval may be divided into n equal subintervals, and therefore G(a, s) is the generating function of
an infinitely divisible distribution. Using the result of Problem (5.12.13b), A.(s) may be written in the
form A.(s) = (A(s) l)A. for some A. and some probability generating function A(s) = L~ aisi. We
now use (iii): if lEI= a,
IP(X (B) :0:: 1)
IP(X(B) = 1)
as a .,), 0. Therefore ao + a 1 = 1, and hence A(s) = ao
distribution with parameter proportional to IB 1.
+ (1 ao)s,
30. (a) Let M (r, s) be the number of points of the resulting process on R+ lying in the interval (r, s].
Since disjoint intervals correspond to disjoint annuli of the plane, the process M has independent
increments in the sense that M(rJ, SJ), M(r2, s2), ... , M(rn, sn) are independent whenever r1 <
SJ < r2 < < rn < Sn. Furthermore, for r <sand k :0:: 0,
{A.:rr(s _
IP(M(r, s) = k) = IP(N has k points in the corresponding annulus) =
r)}keJ..n(sr)
k!
Loo
(A.nx2t eJ..nx2
r= k
r.1
and the claim follows by differentiating, and utilizing the successive cancellation.
31. The number X(S) of points within the sphere with volume Sand centre at the origin has the
Poisson distribution with parameter A.S. Hence IP(X (S) = 0) = eJ..S, implying that the volume V of
the largest such empty ball has the exponential distribution with parameter A..
315
[6.15.32][6.15.33]
Solutions
Markov chains
eJ...crn
r 2: 0.
32. The time between the kth and (k + 1)th infection has mean 'Ak" 1, whence
N
E(T)
=I:.
Ak
k=1
Now
N
{N
k + .(.; N
1
}
+ 1 k
1
2
L= {logN +y +O(N 1)}.
N +1
k
N +1
2
= 
k= 1
It may be shown with more work (as in the solution to Problem (5.12.34)) that the moment
generating function of 'A(N + l)T 2log N converges as N+ oo, the limit being {r(l 8)} 2 .
33. (a) The forward equations for Pn (t) = JP'(V (t) = n + ~) are
p~(t)
n 2: 0,
( 2s+G
aa
)
as
as
( s 2 +sG
aa
)
as
t) =
 1J
1s
1 )
(t + 1s
for some function f. The boundary condition is G(s, 0) = 1, and the solution is as given.
(b) Clearly
mn(T)=E(foT lntdt) =loT ECint)dt
by Fubini's theorem, where lnt is the indicator function of the event that V (t) = n + ~.
As for the second part,
Loo snmn(T) =
n=O
r G(s,
lo
T
t) dt = log[ 1 + (1  s)T],
1 s
Loo
n=O
1
( 1 + (1 s)T)
log(1 s)
sn (mn(T) log T) =log
+
=
1s
316
1s
Loo snan
n=1
Solutions [6.15.34][6.15.35]
Problems
(0, 1)
(1, 0)
34. It is clear that Y is a Markov chain, and its possible transitions are illustrated in the above
diagram. Let x andy be the probabilities of ever reaching (1, 1) from (1, 2) and (2, 1), respectively.
By conditioning on the first step and using the translational symmetry, we see that x = ~ y + ~ x 2
andy= ~ + ~xy. Hence x 3  4x 2 + 4x 1 = 0, an equation with roots x = 1, ~(3 0). Since
xis a probability, it must be that either x = 1 or x = ~(3 0), with the corresponding values of
y = 1 and y = ~ (0  1). Starting from any state to the right of (1 , 1) in the above diagram, we
see by recursion that the chance of ever visiting (1, 1) is of the form xa yf3 for some nonnegative
integers a, {3. The minimal nonnegative solution is therefore achieved when x = ~(3 0) and
y = ~(0 1). Since x < 1, the chain is transient.
35. We write A, 1, 2, 3, 4, 5 for the vertices of the hexagon in clockwise order. Let Ti = min{n :::
1: Xn = i} and ll"iO = ll"( I Xo = i).
(a) By symmetry, the probabilities Pi = ll"i (h < Tc) satisfy
2
PA = 3PI
whence p A =
I
PI= 3
I
+ 3P2
I
P2 = 3PI
1
+ 3P3,
P3 = 3P2,
:J:r.
:!, JTi
ftAJTC
= 2.
(d) We condition on the event E = {h < Tc} as in the solution to Exercise (6.2.7). The probabilities
bi = lP'i (E) satisfy
317
[6.15.36][6.15.37]
Solutions
yielding bi = ft, b2 =
equations of the form
j;, b3
Markov chains
ii2 =
li"2(E)p12
li"I (E)
giving ILIA=
~IL3A,
WI'th
IL3A = 1 + IL2A
.
.
theobvwus
notation,
ILIA= 1 + +IL2A
Pi,i+I =
m2
a(i
Pi+I,i =
Pi,i+I =
m
Look for a solution to the detailed balance equations
f3(m  i)
lri
= lri+I
a(i
+ 1)
m
+ 1)
m
by the concavity of c
318
Solutions [6.15.38][6.15.41]
Problems
where we have used the fact that 'E,j njPjk(t) = nk. Now aj(s)+ nj ass+ oo, and therefore
d(t) + c(l).
I X(O) = 0)
'
(u(t))
= "'no
L..J :1P'(X(2t) = j I X(t) = O)uj(t) =no "
L..Jnj
 1 .
j
n]
n]
The function c(x) =  x 2 is concave, and the claim follows by the result of the previous problem.
39. This may be done in a variety of ways, by breaking up the distribution of a typical displacement and
using the superposition theorem (6.13.5), by the colouring theorem (6.13.14), or by Renyi's theorem
(6.13.17) as follows. Let B be a closed bounded region of Rd. We colour a point of IT at x E Rd
black with probability lP'(x + X E B), where X is a typical displacement. By the colouring theorem,
the number of black points has a Poisson distribution with parameter
{
A.lP'(x + X E B) dx = A. {
}Jf!.d
=A.
lP'(X E dy  x)
dy {
}yEB
lxEJRd
dy
}yEB
lP'(X
dv) = A.IBI,
lvE!Rd
by the change of variables v = y  x. Therefore the probability that no displaced point lies in B is
eJ..!BI, and the claim follows by Renyi's theorem.
40. Conditional on the number N (s) of points originally in the interval (0, s ), the positions of these
points are jointly distributed as uniform random variables, so the mean number of these points which
lie in ( oo, a) after the perturbation satisfies
A.s
looo Fx(au)du=IE(RL)
lo0s1lP'(X+u:sa)du+A.
s
0
as s + oo,
where X is a typical displacement. Likewise, IE(LR) =A. J000 [1  Fx(a + u)] du. Equality is valid
if and only if
00
[1 Fx(v)]dv
jaoo
Fx(v)dv,
41. Conditional on the number N(t) of arrivals by timet, the arrival times of these ants are distributed
as independent random variables with the uniform distribution. Let U be a typical arrival time, so
that U is uniformly distributed on (0, t). The arriving ant is in the pantry at time t with probability
n = lP'(U +X > t), or in the sink with probability p = lP'(U +X < t < U +X+ Y), or departed
with probability 1  p  n. Thus,
IE(xA(t)YB(t)) = IE{IE(xA(t)YB(t) I N(t))}
= IE{ (n x + py + 1 
n  p )N(t)}
= eh(x1) eJ..p(y1).
Thus A(t) and B(t) are independent Poissondistributed random variables. If the ants arrive in pairs
and then separate,
319
[6.15.42][6.15.45]
Solutions
Markov chains
where y = 1 n p. Hence,
42. The sequence {Xr} generates a Poisson process N(t) = max{n : Sn _:::: t}. The statement that
Sn =tis equivalent to saying that there are n 1 arrivals in (0, t), and in addition an arrival at t. By
Theorem (6.12.7) or Theorem (6.13.11), the first n  1 arrival times have the required distribution.
Part (b) follows similarly, on noting that fu(u) depends on u = (UJ, u2, ... , un) only through
the constraints on the Ur.
43. Let Y be a Markov chain independent of X, having the same transition matrix and such that Yo
has the stationary distribution Jr. LetT= rnin{n ::: 1 : Xn = Yn} and suppose Xo = i. As in the
proof of Theorem (6.4.17),
IPij (n) Tl:j I =
IL
I_: L
Now,
lP'(T > r
where
+1 IT
for r::: 0,
> r) _:::: 1 E2
E2
= n12
+ n; (a; (s) 
1
_:: : 2 LL(;..s+t +"At +"As +"Ar +"Am)
n r m
as n+ oo,
where 0 < A < oo. For the last part, use the fact that I:~;:J f(Xr) = l:::iES f(i)V; (n). The result
is obtained by Minkowski's inequality (Problem (4.14.27b)) and the first part.
45. We have by the Markov property that f(Xn+i I Xn, Xni, ... , Xo) = f(Xn+i I Xn), whence
E(log f(Xn+i I Xn, XnJ, ... , Xo) I Xn, ... , Xo) = E(log f(Xn+i I Xn) I Xn).
320
Problems
Solutions
[6.15.46][ 6.15.48]
JC,
46. Let T = inf{t : X 1 = Y1 }. Since X and Y are persistent, and since each process moves by
distance 1 at continuously distributed times, it is the case that lP'(T < oo) = 1. We define
Zt= {
Xt
if t < T,
Yt
ift 2: T,
+ lP'(Yt
= k, T > t).
and therefore
E(sX(t)) = E(sU )E(s V) = (set
+1
sr (s1flt
+1
s)nr.
Also, E(X(t)) = re 1 + (n r)1/1 1 and var(X(t)) = re 1 (1 e1 ) + (n r)1/f1 (1 1/11 ). In the limit as
n + oo, the distribution of X (t) approaches the bin(n, A./(A. + ft)) distribution.
48. Solving the equations
gives the first claim. We have that y = l:i (Pi  qi )Jri, and the formula for y follows.
Considering the three walks in order, we have that:
A. Jri =
B. Substitution in the formula for YB gives the numerator as 3{ ~a +o(a) }, which is negative
for small a whereas the denominator is positive.
321
[6.15.49][6.15.51]
Solutions
Markov chains
i (Jb 
C. The transition probabilities are the averages of those for A and B, namely, Po =
a)+
a) =
a, and so on. The numerator in the formula for YC equals
+o(l),
which is positive for small a.
i<i
fu
rfu
49. Call a car green if it satisfies the given condition. The chance that a green car arrives on the scene
during the time interval (u, u +h) is A.hlP'(V < xf(t u)) for u < t. Therefore, the arrival process
of green cars is an inhomogeneous Poisson process with rate function
A.(u) = {
if u < t,
if u :::: t.
Hence the required number has the Poisson distribution with mean
A.
50. The answer is the probability of exactly one arrival in the interval (s, t), which equals g(s) =
A.(t s)eJ..(ts). By differentiation, g has its maximum atS = max{O, t  A. 1}, and g(s) = e 1
whent:::: A. 1 .
51. We measure money in millions and time in hours. The number of available houses has the
Poisson distribution with parameter 30A., whence the number A of affordable houses has the Poisson
distribution with parameter~ 30A. =SA. (cf. Exercise (3.5.2)). Since each viewing timeT has moment
generating function E(e 8T) = (e 28  e8 )je, the answer is
322
7
Convergence of random variables
2.
(a) E({aX
+ bY}Z) =
+ E({X 
(b) E({X + Y} 2 )
(c) Clearly
aE(XZ)
+ bE(YZ).
+ 2E(Y 2 ).
Y} 2 ) = 2E(X 2 )
3. Let f(u) = ~E, g(u) = 0, h(u) = ~E, for all u. Then df(f, g)+ dE(g, h) = 0 whereas
df(f, h)= 1.
~
Either argue directly, or as follows. With any distribution function F, we may associate a graph
F obtained by adding to the graph of F vertical line segments connecting the two endpoints at each
discontinuity ofF. By drawing a picture, you may see that .../2 d (F, G) equals the maximum distance
between F and G measured along lines of slope 1. It is now clear that d(F, G) = 0 if and only if
F = G, and that d(F, G) = d(G, F). Finally, by the triangle inequality for real numbers, we have
that d(F, H) ~ d(F, G)+ d(G, H).
5.
Take X to be any random variable satisfying E(X 2 ) = oo, and define Xn = X for all n.
let n + oo to obtain lim infn+oo EIX~ I 2: EIX' 1. By another application ofMinkowski's inequality,
323
[7.2.2][7.2.4]
Solutions
i,
(c) By part (a), E(X~) + E(X 2 ). Now Xn _..;. X by Theorem (7.2.3), and therefore E(Xn) + E(X)
by part (b). Therefore var(Xn) = E(X~)  E(Xn) 2 + var(X).
2. Assume that Xn ~ X. Since IXnl ~ Z for all n, it is the case that lXI
Zn = IXn XI satisfies Zn ~ 2Z a.s. In addition, if E > 0,
~ Z a.s. Therefore
As n+ oo, lP'(IZnl >E) + 0, and therefore the last term tends to 0; to see this, use the fact that
E(Z) < oo, together with the result of Exercise (5.6.5). Now let E ,!.. 0 to obtain that EIZnl + 0 as
n + oo.
3. We have that X n 1 ~ Xn ~ X, so that E(Xn) + E(X), and similarly E(Yn) + E(Y). By
the independence of Xn and Yn,
{( 1) ( 1) }
X  ;;_
Y  ;;_
= E(XY) 
E(X) + E(Y)
1
n
+ n 2 + E(XY)
4. Let F1, F2, ... be distribution functions. As in Section 5.9, we write Fn + F if Fn(x) + F(x)
for all x at which F is continuous. We are required to prove that Fn + F if and only if d (Fn, F) + 0.
Suppose that d(Fn, F) + 0. Then, forE > 0, there exists N such that
F(x E) E ~ Fn(x)
F(x +E)+ E
for all x.
Take the limits as n + oo and E + 0 in that order, to find that Fn (x) + F (x) whenever F is
continuous at x.
Suppose that Fn + F. Let E > 0, and find real numbers a = x1 < x2 < < Xn = b, each
being points of continuity of F, such that
(i) Fi(a) < E for all i, F(b) > 1 E,
(ii) lxi+1 Xi I < E for 1 ~ i < n.
In order to pick a such that Fi (a) < E for all i, first choose a 1 such that F(a 1) < iE and F is
continuous at a', then find M such that IFm (a') F(a 1 ) I < iE form ::::: M, and lastly find a continuity
point a ofF such that a~ a 1 and Fm(a) < E for 1 ~ m < M.
There are finitely many points Xi, and therefore there exists N such that IFm(Xi) F(xi)l < E
for all i and m ::::: N. Now, if m ::::: N and Xi ~ x < Xi+ 1,
and similarly
Fm(x)::::: Fm(Xi) > F(xi) E :0:: F(x E) E.
324
Modes of convergence
5.
n:::::
Solutions [7.2.5][7.2.7]
(a) Suppose c > 0 and pick 8 such that 0 < 8 < c. Find N such that lP'(/Yn  cJ > 8) < 8 for
N. Now, for x::::: 0,
8) +
IP'(/Yn c/ >
8)
and similarly
8)
Taking the limits as n + oo and 8 .,),. 0, we find that lP'(XnYn .::; x) + lP'(X .::; xfc) if xfc is a point
of continuity of the distribution function of X. A similar argument holds if x < 0, and we conclude
that XnYn
For the second part, it suffices to prove that Y,; 1 ~ c 1 if Yn ~ c (# 0). This is immediate
from the fact that IYn 1  c 1 1 < E/{/c/(/c/ E)} if /Yn  c/ < E ( < Jcl).
(b) Let E > 0. There exists N such that
lP'(/Xn/ >E)< E,
IP'(/Yn Y/ >E)< E,
ifn
::0:: N,
and in addition lP'(/YI > N) <E. By an elementary argument, g is uniformly continuous at points of
the form (0, y) for Jy/ .::; N. Therefore there exists 8 (> 0) such that
y/ .::; 8.
If /Xn/.::; 8, /Yn Y/ .::; 8, and IYI .::; N, then Jg(Xn, Yn) g(O, Y)J < E, so that
::0:: E).::;
6.
p
+ g(O,
Y) as n+ oo.
nU n
00
A=
00
00
k=1 n=1m=1
X(w) = {
limn+oo Xn(w)
ifw E A
ifw
rf. A.
IP'(/cnXn eX/> E).::; IP'(cn/Xn X/> iE) + IP'(/cn cJJXJ > iE)
.:'S lP' (/Xn X/ >
+ 0
E
) + lP' (/X/ >
E
)
2(c +E)
2/cn  cJ
as n + oo.
325
[7.2.8][7.3.1]
Solutions
8.
Y. Then cnYn
If X is not a.s. constant, there exist real numbers c and E such that 0 < E <
lP'(Xn<c)>E,
lP'(Xn>c+E)>E,
ifn;:::N.
Also, by the triangle inequality, IXr Xsl _:::: IXr XI+ IXs XI; therefore there exists M such
that IP'(IX,  Xs I > E) < E3 for r, s ::::: M. Assume now that the Xn are independent. Then, for
r, s ::::: max{M, N}, r =f. s,
E3 > IP'(IXr  Xs I > E)
IP'(Xr < c, Xs > c +E) = lP'(Xr < c)lP'(Xs > c +E) > E2 ,
::0::
a contradiction.
9. Either use the fact (Exercise (4.12.3)) that convergence in total variation implies convergence in
distribution, together with Theorem (7.2.19), or argue directly thus. Since luOI :::: K < oo,
IE(u(Xn)) E(u(X))I = jL:u(k){fn(k) /(k)}l _:::: KL lfn(k) /(k)l+ 0.
k
= 2::~= 1
= 2::~= 1 A.,.
For fixed
x, the event {Sn _:::: x} is decreasing inn, whence by Lemma (1.3.5), if an + a < oo and xis a
nonnegative integer,
oo
lP' ( LXr
r=1
crj
)=0
L~
Hence if a < oo, 2::~ 1 X, converges to a Poisson random variable. On the other hand, if an + oo,
L:J=
then ecrn
0 aj I j! + 0, giving that IP'(2::~ 1 X, > x)
diverges with probability 1, as required.
I for all
x,
(a) If IXn  Xm I > E then either IXn XI > iE or IXm XI > iE, so that
m::::: nk.
The sequence (nk) may not be increasing, and we work instead with the sequence defined by N1 = n1,
Nk+1 = max{Nk + 1, nk+d We have that
LlP'(IXNk+J XNkl::::: Tk) <
k
326
00,
Solutions
[7.3.2][7.3.3]
whence, by the first BorelCantelli lemma, a.s. only finitely many of the events {IXNk+l  XNk I 2::
2k} occur. Therefore, the expression
00
X= XN 1 + L(XNk+i  XNk)
k=1
converges absolutely on an event C having probability one. Define X(w) accordingly for wE C, and
X(w) = 0 for w rf. C. We have, by the definition of X, that XNk ~ X ask+ oo. Finally, we 'fill
in the gaps'. As before, forE > 0,
as n, k+ oo, where we are using the assumption that {Xn} is Cauchy convergent in probability.
(b) Since Xn
as n, m+ oo,
forE > 0. Therefore {Yn} is Cauchy convergent also, and the sequence converges in probability to
some limit Y. Finally, Xn
2.
lJ
lJ
e:)
{p(l  p W, so that
LlP'(Szn = 0) < oo
if p
=f.
i.
,Jfi,) =
1  <I>(l) > 0,
where <I> is theN (0, 1) distribution function. Since {J 2:: 1} is a tail event of an independent sequence,
it has probability either 0 or 1, and therefore lP'(/ _:::: 1) = lP'(J 2:: 1) = 1, and also lP'(/ _:::: 1, J 2::
1) = 1. That is, on an event having probability one, each visit of the walk to the left of .[ii is
followed by a visit of the walk to the right of .[ii, and vice versa. It follows that the walk visits 0
infinitely often, with probability one.
327
[7.3.4][7.3.8]
Solutions
4. Let A be exchangeable. Since A is defined in terms of the Xi, it follows by a standard result of
measure theory that, foreachn, thereexistsaneventAn E a(X1, X2, ... , Xn), suchthatlP'(Ab.An)+
0 as n + oo. We may express An and A in the form
An = {Xn E Bn},
where Xn
A = {X E B},
= (X 1, X2, ... , Xn), and Bn and Bare appropriate subsets ofll~n and JR 00 Let
A~ ={X~
Bn},
A'= {X 1 E B},
where X~ = (Xn+1, Xn+2, ... , X2n) and X'= (Xn+1, Xn+2, ... , X2n, X1, X2, ... , Xn, X2n+1,
X2n+2, ).
Now !P'(An n A~) = lP'(An)lP'(A~), by independence. Also, !P'(An) = lP'(A~), and therefore
!P'(A
By the exchangeability of A, we have that lP'(A t. A~) = !P'(A' t. A~), which in turn equals
t. An), using the fact that the Xi are independent and identically distributed. Therefore,
liP'( An
n A~)
t.
An)
as n + oo.
Combining this with(*), we obtain that lP'(A) = lP'(A) 2, and hence lP'(A) equals 0 or 1.
5. The value of Sn does not depend on the order of the first n steps, but only on their sum. If Sn = 0
i.o., then S~ = 0 i.o. for all walks {S~} obtained from {Sn} by permutations of finitely many steps.
6. Since f is continuous on a closed interval, it is bounded: If (y) I ::: c for all y E [0, 1]for some c.
Furthermore f is uniformly continuous on [0, 1], which is to say that, if E > 0, there exists 8 (> 0),
such that lf(y) f(z)l < E if IY zl :S 8. With this choice of E, 8, we have that IE(Z/Ac)l < E, and
IE(ZIA)I ::: 2clP'(A) ::: 2c
x(1  x)
n 82
7. If {Xn} converges completely to X then, by the first BorelCantelli lemma, IXn XI > E only
finitely often with probability one, for all E > 0. This implies that Xn ~ X; see Theorem (7.2.4c).
Suppose conversely that {Xn} is a sequence of independent variables which converges almost
surely to X. By Exercise (7.2.8), X is almost surely constant, and we may therefore suppose that
Xn ~ c where c E JR. It follows that, for E > 0, only finitely many of the (independent) events
{IXn  cl > E} occur, with probability one. Using the second BorelCantelli lemma,
LlP'(IXn cl >E)<
00.
8.
(n)
2
1
_n (_!_ ~
1 n L
z"1
328
xi)2
1
x2
n(n  1) L. 1
z1
Solutions [7.3.9][7.3.12]
Now n
1
L;'i Xi
P
n 1 L;'i Xi +
~t (use Theorem
(7.2.4a)). It follows that (n 1 L;'i Xi) 2 ~ ~t 2 ; to see this, either argue directly or use Problem
(7.11.3). Now use Exercise (7.2.7) to find that
n ( 1 n
I: xi )2 +p ~t 2 .
n 1 n i=
1
Arguing similarly,
   """'x? ~ o
n(n 1) ~
!=1
'
'
and the result follows by the fact (Theorem (7.3.9)) that the sum of these two expressions converges
in probability to the sum of their limits.
9.
Evidently,
Xn
) = 1lP' ( >1+E
logn nl+E'
10. (a) Mills's ratio (Exercise (4.4.8) or Problem (4.14.1c)) informs us that 1 <P(x) ~ x 1cp(x) as
x + oo. Therefore,
lP'(IXnl ~ y'2logn(l+E)) ~
2.
The sum over n of these terms converges if and only if E > 0, and the BorelCantelli lemmas imply
the claim.
(b) This is an easy implication of the BorelCantelli lemmas.
11. Let X be uniformly distributed on the interval [1, 1], and define Xn = /{X:S(l)n /n) The distribution of Xn approaches the Bernoulli distribution which takes the values 1 with equal probability
The median of X n is 1 if n is even and  1 if n is odd.
i.
= 0) = 1.
329
1.
= oo.
[7 .3.13][7.4.3]
Solutions
13. We divide the numerator and denominator by Jna. By the central limit theorem, the former
converges in distribution to the N(O, 1) distribution. We expand the new denominator, squared, as
12:n
na 2
(Xr 
~t)
2
 (Xna 2
r=1
~t)
I:n (Xr
~t)
1
+(Xa2
r=1
~t)
By the weak law of large numbers (Theorem (5.10.2), combined with Theorem (7.2.3)), the first term
converges in probability to 1, and the other terms to 0. Their sum converges to 1, by Theorem (7.3.9),
and the result follows by Slutsky's theorem, Exercise (7.2.5).
LetSn=X1+X2++Xn.Then
n
'"'L
1 <
nE(S 2) =
n
i= 2 log i  log n
and therefore Snfn ~ 0. On the other hand, I::i lP'(IXil 2: i) = 1, so that IXil 2: i i.o., with
probability one, by the second BorelCantelli lemma. For such a value of i, we have that ISi  Si 11 2:
i, implying that Sn/ n does not converge, with probability one.
2.
1
n
lP'(Xn = n) = 1  2
lP'(Xn
= n3 
n)
= 21 ,
n
implying by the first BorelCantelli lemma that lP'(Xnfn+ 1) = 1. It is an elementary result of
real analysis that n  1 I::~= 1 Xn + 1 if Xn + 1, and the claim follows.
3. The random variable N(S) has mean and variance AlSI = crd, where cis a constant depending
only on d. By Chebyshov's inequality,
lP'
(I
I )
N(S)
ISlA :::
A (A)
~
.:5 E21SI =
era
where Skis the sphere of radius k. It follows that N(Sk)/ISkl ~ A ask+ oo. The same conclusion
holds ask + oo through the reals, since N(S) is nondecreasing in the radius of S.
330
Solutions
Martingales
[7.5.1][7.7.1]
Let Iij be the indicator function of the event that X J lies in the i th interval. Then
n
log Rm
=L
Zm (i) log Pi
=L
i=l
lij log Pi
=L
i=l }=I
l:t=I Iij
Yj
}=I
lE(Yj) =
L Pi log Pi = h.
i=l
as t+ oo,
By the strong law, Sn/n ~ lE(XJ) =I= 0. In particular, with probability 1, Sn = 0 only finitely
often.
3.
~
alogn
for all large n, by the tail estimate of Exercise (4.4.8) or Problem (4.14.1c) for the normal distribution.
This is summable if a > 2, and the claim follows by an application of the first BorelCantelli lemma.
_I]}
= lE{ Xi [lE(SJ I So, S1, ... , SJI) s1_I]} = 0
331
[7. 7.2][7.8.2]
Solutions
1 {
( 1 _ p,n+1)}
E(Sn+1 I Zo, z1' ... ' Zn) = p,n+1 E(Zn+1 I Zo, ... ' Zn) m
1  JL
1 {
( 1 _ p,n+1)}
= p,n+ 1 m+p,Znm
=Sn.
1 _p,
3.
+ Xn
where we have used the fact that Zn+1 depends only on X1, X2, ... , Xn.
E(S~) = I:var(Xj).
}=1
2. It would be easy but somewhat perverse to use the martingale convergence theorem, and so we
give a direct proof based on Kolmogorov 's inequality of Exercise (7 .8 .1 ). Applying this inequality to
the sequence Zm, Zm+1' ... where Zi =(Xi EXi)/i, we obtain that Sn = Z1 + Z2 + + Zn
satisfies, for E > 0,
lP' ( max ISn Sml
m<n<r
 
~
~
var(Zn).
n=m+1
1 00
1
lP' ( sup ISn Sml > E ) ::0 2
2 var(Xn).
n:=::m
E n=m+1 n
.2:::
E) = 1
for all
m;.oo n:=::m
Xm
I ::0
332
for all
> 0,
> 0.
Solutions
[7.8.3][7.9.4]
is Cauchy convergent, and hence convergent. It follows that Sn converges a.s. to some limit Y.
The last part is an immediate consequence, using Kronecker's lemma.
3. By the martingale convergence theorem, S = limn+oo Sn exists a.s., and Sn
Exercise (7.2.1c), var(Sn) + var(S), and therefore var(S) = 0.
m.s.
S. Using
I Y) = Y 2 , lE(Y I X) = 0.
since lE(Y) = lE(Y 3 ) = 0. This is a minimum when b = JE(Y 2 ) = ~.and a = 0. The best linear
predictor of X given Y is therefore ~ .
Note that lE(Y I X) = 0 is a linear function of X; it is therefore the best linear predictor of Y
given X.
2.
3.
Write
~tz
+ paz(X 
~tdfal,
and
v(a) = lE{ (Y g(a)) 2 } = lE(Y 2 )  2alE(YX1) + aVa'.
Let a be a vector satisfying Vi' = lE(YX'). Then
v(a)  v(a) = aVa'  2alE(YX1) + 2ilE(YX1) iVa'
= aVa' 2aVa' +iVa'= (a a)V(a a)' :::: 0,
since Vis nonnegative definite. Hence v(a) is a minimum when a =a, and the answer is g(i). If V
is nonsingular, a= JE(YX)V 1 .
4. Recall that Z = JE(Y I fj,) is the ('almost') unique fJ,measurable random variable with finite
mean and satisfying JE{(Y Z)/G} = 0 for all G E fJ,.
333
[7.9.5][7.10.1]
Solutions
(vi) If g is convex then, for all a E JR, there exists A.(a) such that
g(y) 2: g(a)
+ (y a)A.(a);
which tends to 0 as m + oo, by the dominated convergence theorem. Therefore E(V) = 0, and hence
lP'(V = 0) = 1. The claim follows.
5.
E(Y
I X)=
X.
6. (a) Let {Xn : n ::: 1} be a sequence of members of H which is Cauchy convergent in mean
square, that is, E{IXn  Xm 12 } + 0 as m, n + oo. By Chebyshov's inequality, {Xn : n ::: 1} is
Cauchy convergent in probability, and therefore converges in probability to some limit X (see Exercise
(7.3.1)). It follows that there exists a subsequence {Xnk : k ::: 1} which converges to X almost surely.
Since each Xnk is fJ,measurable, we may assume that X is fJ,measurable. Now, as n + oo,
where we have used Fatou's lemma and Cauchy convergence in mean square. Therefore Xn ~ X.
That E(X 2 ) < oo is a consequence of Exercise (7.2.la).
(b) That (i) implies (ii) is obvious, since I G E H. Suppose that (ii) holds. Any Z ( E H) may be
written as the limit, as n + oo, of random variables of the form
m(n)
Zn =
a;(n)Ic;(n)
i=l
for reals a;(n) and events G;(n) in fJ,; furthermore we may assume that IZnl :<:: IZI. It is easy to see
that E{ (Y M)Zn} = 0 for all n. By dominated convergence, E{(Y M)Zn} + E{(Y M)Z},
and the claim follows.
It is easily checked by considering whether lxl :<::a or IYI :<::a that, for a > 0,
334
Solutions
Uniform integrability
[7.10.2][7.10.6]
xn
2. (a) Let E > 0. There exists N such that lE(IXn < E if n > N. Now lEIXrl < oo, by
Exercise (7.2.1a), and therefore there exists 8 (> 0) such that
lE(IXnlr lA) < E for 1 S n S N,
+ {lE(IX(/A)} 1/r
S 2El/r
ifn > N
liminflEIX~kl
S suplEIX~I
k+oo
n
< oo.
lE(IX~IIA) < E
for all n,
whenever A is such that lP'(A) < 8. There exists N such that Bn(E) = {IXn XI > E} satisfies
lP'(Bn(E)) < 8 for n > N. Consequently
lE(IXn
xn S
Er
n > N,
S 2El/r.
Therefore, Xn ~X.
3.
FixE > 0, and find a real number a such that g(x) > xjE if x >a. If b 2: a,
lE (IXnii(IXnl>bJ) < ElE{g(IXnl)} S EsupE{g(IXnl)},
n
4.
Here is a quick way. Extinction is (almost) certain for such a branching process, so that Zn ~ 0,
and hence Zn ~ 0. If {Zn : n 2: 0} were uniformly integrable, it would follow that JE(Zn) + 0 as
n + oo; however lE(Zn) = 1 for all n.
5.
We may suppose that Xn, Yn, and Zn have finite means, for all n. We have that 0 S Yn  Xn S
p
= lEIZ XI,
6. For any event A, lE(IXniiA) S JE(ZIA) where Z = supn IXnl The uniform integrability follows
by the assumption that lE(Z) < oo.
335
[7.11.1][7.11.2]
Solutions
as n+ oo,
T(
so that Xn + 0.
You have insufficient information to decide whether or not Xn converges almost surely:
(a) Let X be Cauchy, and let Xn = Xjn. Then Xn has the required density function, and Xn ~ 0.
(b) Let the Xn be independent with the specified density functions. ForE > 0,
2 sin 1 (
lP'(IXnl >E)=n
,h + n2E2
) "'2,
nnE
so that Ln lP'(IXnl >E)= oo. By the second BorelCantelli lemma, IXnl > E i.o. with probability
one, implying that Xn does not converge a.s. to 0.
2. (i) Assume all the random variables are defined on the same probability space; otherwise it is
meaningless to add them together.
(a) Clearly Xn(w) + Yn(w) + X(w) + Y(w) whenever Xn(w) + X(w) and Yn(w) + Y(w).
Therefore
{ Xn + Yn...,. X+ Y} ~ {Xn...,. X} U {Yn...,. Y},
a union of events having zero probability.
(b) Use Minkowski's inequality to obtain that
{lE (IXn + Yn X Yn} 1/r .::0 {lE(IXn 
xn} 1/r +
nl >E)
336
8),
Solutions
Problems
[7 .11.3][7.11.5]
which tends to 0 in the limit as n + oo and 8 + oo in that order. Together with two similar facts, we
obtain that XnYn ~ XY.
(h) The example of (d) above indicates that the corresponding statement is false for convergence in
distribution.
3. Let E > 0. We may pick M such that lP'(IXI 2: M) .::::E. The continuous function g is uniformly
continuous on the bounded interval [ M, M]. There exists 8 > 0 such that
lg(x) g(y)l
If lg(Xn) g(X)I >
E,
.:::= E
.:::=
8 and lxl
.:::=
M.
lP'(Ig(Xn) g(X)I
4.
if lx Yl
+ JI(IXI 2:
M)
+ lP'(IXI 2:
M) _:: : E,
g(X).
Clearly
lE(eitXn)
n
.
n { 1 1
=II
lE(eitYj/101) =II _.  e .
.
.
.
10
1 _ eztf10J
it/10jl }
]=1
]=1
1 eit
wn(leitjlOn)
1  eit
+
it
as n + oo. The limit is the characteristic function of the uniform distribution on [0, 1].
Now Xn .::; Xn+1 _:::: 1 for all n, so that Y(w) = limn+oo Xn(w) exists for all w. Therefore
Xn
5.
Y; hence Xn
lP'(IN(t+h)N(t)l
(I
N(t +
337
1)
= A.h + o(h),
[7.11.6][7.11.10]
Solutions
6.
By Markov's inequality, Sn
= 2:7= 1 X;
satisfies
since E(X;) = 0 for all i. Therefore there exists a constant C such that
7.
00
8. Either use the Skorokhod representation or characteristic functions. Following the latter route,
the characteristic function of aXn + b is
where >n is the characteristic function of Xn. The result follows by the continuity theorem.
9.
= li"(X + c > t + c)
<

E{(X+c) 2 }
.
(t +c) 2
10. Note that g(u) = uj(l + u) is an increasing function on [0, oo). Therefore, forE > 0,
lP' X >
(I nl
=11" (
IXnl
l+E
> E ) <  .JE: ( IXnl )
l+IXnl
l+E E
l+IXnl
338
Solutions [7.11.11][7.11.14]
Problems
~ 0.
IXnl )
E
E
:'::   li"(IXnl :'::E)+ 1 lP'(IXnl >E)+1+1Xnl
1+E
1+E
as n + oo, for E > 0. However E is arbitrary, and hence the expectation has limit 0.
11. (i) The argument of the solution to Exercise (7.9.6a) shows that {Xn} converges in mean square
if it is meansquare Cauchy convergent. Conversely, suppose that Xn ~ X. By Minkowski's
inequality,
xi is
1 n
2 L:var(Xi) :=:+ 0
n i=1
n
as n+ oo.
= 0.
as n+
00.
n+oo
since y 1 log(1  y) + 1 as y t 0. Setting x = 1, we obtain n(l  F(an)) + log H(l), and
the second limit follows.
(b) This is immediate from the fact that it is valid for all sequences {an}.
(c) We have that
1 F(tex+y)
1  F(t)
1 F(tex+y)
1 F(tex)
1F(tex)
1  F(t)
logH(eY)
logH(ex)
log H(l)
as t + oo. Therefore g(x + y) = g(x)g(y). Now g is nonincreasing with g(O) = 1. Therefore
g(x) = ef3x for some ,B, and hence H(u) = exp( auf3) for u > 0, where a= log H(l).
14. Either use the result of Problem (7.11.13) or do the calculations directly thus. We have that
JP>(Mn :=oxn/rr)
= {~ +~tan 1
c:) r=
{1
~tan 1 (;J
339
[7.11.15][7.11.16]
Solutions
if /lu/1 00 :::0 1. There is equality if Un equals the sign of fn  gn. The second equality holds as in
Problem (2.7.13) and Exercise (4.12.3).
(b) Similarly, if /lu/1 00 :::0 I,
IE(u(X)) E(u(Y))I :::0
L:
L:
lf(x) g(x)l dx
with equality if u(x) is the sign of f(x) g(x). Secondly, we have that
llP'(X
A) lP'(Y
A) I
where
u(x) {
if X E A,
1
ifx.A.
z:dTV(X, Y) = lP'(X
=f
Y) =
1 if JL = 0,
ifJL=l,
and therefore we may assume that 0 < JL < I. Let U, V, W be random variables with mass functions
lP'(u _

Xn
) _ J.Ln
JL
lP'(V _
'
Xn
) _  min{fn gn, 0}
Xn
1J.L
'
and let Z be a Bernoulli variable with parameter JL, independent of (U, V, W). We now choose the
pair X', Y' by
(X' Y') = { (U, U) if Z = I,
'
(V,W) ifZ=O.
340
Solutions [7.11.17][7.11.19]
Problems
It may be checked that X' andY' have the same distributions as X andY, and furthermore, li"(X'
Y 1) = li"(Z = 0) = 1 fi = idTV(X, Y).
i=
(e) By part (d), we may find independent pairs (X;, Y[), 1 :::0 i :::0 n, having the same marginals as
(Xi, Yi), respectively, and such thatli"(X;
:::: 211" (
i=
L x; i= L:rf
i=l
n
::::2 Lll"(x;
i=l
i= Y[)
i=l
= 2
LdTV(Xj, Yj).
i=l
17. If XJ, X2, ... are independent variables having the Poisson distribution with parameter A, then
Sn = X1 +X2 + +Xn has the Poisson distribution with parameter An. Now n 1sn SA, so that
IE(g(n 1Sn)) + g(A) for all bounded continuous g. The result follows.
Umn=
vm+n
satisfies
log 1/lmn (t) = n (eit!.../m+n 
1)
+ m (eit!.../m+n 
1)
+ (m  n)it + _lt2
2
Jm+n
as m, n+ oo, implying by the continuity theorem that Umn S N(O, 1). Now Xn + Ym is Poissondistributed with parameter m + n, and therefore
_
v:mn
J+
Xn Ym P 1
+
m+n
asm,n+ oo
D
by the law oflarge numbers and Problem (3). It follows by Slutsky's theorem (7.2.5a) that Umn IVmn +
N(O, 1) as required.
19. (a) The characteristic function of Xn is 1/Jn (t) = exp{i ftnt  ia;t 2 } where fin and a; are
I 2
the mean and variance of Xn. Now, limn+oo 1/Jn(l) exists. However 1/Jn(l) has modulus e'Ian,
and therefore a 2 = limn+oo a; exists. The remaining component eitLnt of 1/Jn (t) converges as
n + oo, say eitLnt + 8(t) as n + oo where (t) lies on the unit circle of the complex plane. Now
I 2 2
1/Jn(t) + 8(t)e'Ia
which is required to be a characteristic function; therefore e is a continuous
function oft. Of the various ways of showing that 8(t) = eitLt for some ft, here is one. The sequence
1/Jn(t) = eitLnt is a sequence of characteristic functions whose limit 8(t) is continuous at t = 0.
Therefore e is a characteristic function. However 1/1n is the characteristic function of the constant fin,
which must converge in distribution as n + oo; it follows that the real sequence {/In} converges to
some limit ft, and 8(t) = eitLt as required.
This proves that 1/Jn(t)+ exp{ifl,t ia 2 t 2 }, and therefore the limit X is N(fl,, a 2).
1 ,
ap)}
341
[7.11.20][7.11.21]
Solutions
in the natural notation. Viewed as a function of s and t, this is the joint characteristic function of a
bivariate normal distribution.
When working in such a context, the technique of using linear combinations of Xn and Yn is
sometimes called the 'CramerWold device'.
Y;
E(Tn fn )
Tn
I:t=l Y;.
I n
2
= 2 l:var(X;) + 2
n i=1
n
2:::
i,J=l
nc
2+ 0.
Xj) ::0
l::Oi<} ::on
Xj )I
::: E if li  jl
cov(X;,
2:::
cov(X;,Xj)+
li}19
l::Oi,J::On
::=:: /.
Now
cov(X;,X1):::2n/c+n 2 Ec,
li}1>1
1::oi,}::On
2/c
n
E(Tn / n ) ::0 
+ EC + EC
as n+ oo.
'f'(t
= 2C
100
2
whence
cp(t) cp(O) = _ [
2c
12
cos(tx) d X
x 2 Iogx

00 1 
cos(tx) dx.
x 2 Iogx
11/t
2
11u
and
!00
t2
2
dx+
dx
Iogx
1/t x2Iogx
'
dx
+0
Therefore
as u+ oo,
Iogx
2
1oo 2 dx
  dx
1uoo x2Iogx
Iogu u x 2
I
< 
if t > 0.
2
= 
ulogu'
as t
u > I.
0.
Now cp is an even function, and hence cp' (0) exists and equals 0. Use the result of Problem (7.ll.l5)
to deduce that n 1 I:'i Xi converges in distribution to 0, and therefore in probability also, since 0 is
constant. The X; do not obey the strong law since they have no mean.
342
Solutions [7.11.22][7.11.24]
Problems
and therefore
12
1~
2p1
Xn =  L)Ui Vi) + n
n i= 1
6
as n
oo,
+
. + eztfx)
. dx = lo1 cos(tjx) dx =It I100 cosy dy
lo 1(eztfx
ltl
Y2
2 0
100
ltl
1 cosy
as t+ 0,
dy = 1 /It I+ o(ltl)
I
11:
2100
1)+ 
11:
du
1+u
1
2
as t
2 =
+
oo.
LlP'(IXkl ~
i whenever mn
+
mn) :"':
L
k=mn
k=1
343
k2
+
as n+ oo,
[7.11.25][7.11.26]
Solutions
Zn =
Ynk
+ Bnmn
k=1
where Bnmn is the sum of n mn independent summands each of which takes the values 1, each
possibility having probability
Furthermore
1 mn
2
2:Ynk ::S mn
Jn k=1
Jn
Finally,
var(Un)
t (2  ~)
k=1
so that
1 n 1
var(Un/ Jn) = 2 2 + 2.
n k=1 k
25. (i) Let c/Jn and cp be the characteristic functions of Xn and X. The characteristic function !frk of
XNk is
00
!frk(t)
whence
00
11/fk(t)
cp(t)l:::::
Let E > 0. We have that ifJ} (t) + cp (t) as j + oo, and hence for any T > 0, there exists J (T) such that
lifJ} (t)  cp(t)l < E if j :::: J (T) and It I ::::: T. Finally, there exists K(T) such that lP'(Nk ::::: J (T)) ::::: E
if k :::: K (T). It follows that
if It I ::::: T and k :::: K(T); therefore !frk(t) + cp(t) ask+ oo.
(ii) Let Yn = SUPm;:::n IXm XI. ForE > 0, n :0:: 1,
lP'(IXNk XI>
+ lP'(Yn
> n)
Now take the limit as n + oo and use the fact that Yn ~ 0.
an ,n
 =
+ 1, n)
ITk ( 1 i=O
"'l
k .)
1. ) < exp (  L..J
 .
n i=O n
344
ask +
00.
Solutions
Problems
L_g
En=
Jnn)
(j
:::
(nj+l _
Jn
j!
nl;~n
n) nlen =en
Jn
[7.11.27][7.11.28]
nl
)
(j l)! ,
j!
I
I 2
g(x)e'1.x dx
En+
.J2ii
 M
x
I 2
e'1.x dx
!oM
.J2ii
1e'J.M
.,=
.J2ii
Also,
where k
= LM JnJ.
<lim
.J2ii 
nn+2en
I
n+ oo {
< .
.J2ii
27. Clearly
E(Rn+1 I Ro, R1, ... , Rn)
+ 2).
Rn
n+2
= Rn +  
Hence
and also 0 ::: Sn ::: 1. Using the martingale convergence theorem, S = limn+oo Sn exists almost
surely and in mean square.
k(t)
L8tJ, m(t)
= ro E3)k(t)l,
n(t)
L(l
+ E3)k(t)J
and let Imn(t) be the indicator function of the event {m(t) ::: M(t) < n(t)}. Since M(t)jt ~ e, we
may find T such that EUmn(t)) > IE fort~ T.
We may approximate SM(t) by the random variable sk(t) as follows. With Aj = {ISj  sk(t) I >
E.Jk(t)},
IP'(AM(t)) ::0 IP'(AM(t) lmn(t)
k(t)1
= 1) + IP'(AM(t) lmn(t) = 0)
n(t)1
:::lP'( U Aj)+lP'( U
j=m(t)
<

Aj)+IP'(lmn(t)=O)
j=k(t)
{k(t)  m(t) }a 2
E2k(t)
+
if t
345
{ n(t) k(t) }a 2
E2k(t)
~
T,
+E
[7.11.29][7.11.32]
Solutions
by Kolmogorov's inequality (Exercise (7.8.1) and Problem (7.11.29)). Send t + oo to find that
as t + oo.
which implies the first claim, since k(t)j(et) + 1 (see Exercise (7.2.7)). The second part follows by
Slutsky's theorem (7.2.5a).
+ (Sn 
Now S'fJAk :::: c 2 IAk; the second term on the right side is 0, by the independence of the X's, and the
third term is nonnegative. The first inequality of the question follows. Summing over k, we obtain
E(S;) :::: c 2 1P'(Mn > c) as required.
forE > 0. Take the limit as m, n + oo to obtain in the usual way that {Sr : r :::: 0} is a.s. Cauchy
convergent, and therefore a.s. convergent, if 2::\"' E(X~) < oo. It is shorter to use the martingale
convergence theorem, noting that Sn is a martingale with uniformly bounded second moments.
(ii) Apply part (i) to the sequence Yk = Xkfbk to deduce that 2::~ 1 Xkfbk converges a.s. The claim
now follows by Kronecker's lemma (see Exercise (7.8.2)).
f Xo IJ PN_ij
lJ
i,j
(b) Clearly Lj Pij = 1 for each i, and we introduce Lagrange multipliers {JL; : i E S} and write
V = A.(P) + Li JLi Lj Pij. Differentiating V with respect to each Pij yields a stationary (maximum)
value when (N;j / Pij) + JLi = 0. Hence Lk N;k =  JLb and
z=E\
346
Solutions [7.11.33][7.11.34]
Problems
excursion from i. By the strong Markov property, the T (r) are independent and identically distributed
with mean J.Li . Furthermore,
V;(n)J
::r
V;(n)
V;(n)
::r
"
T(r)
By the strong law of large numbers and the fact that V; (n) ~ oo as n + oo, the two outer terms
sandwich the central term, and the result follows.
(b) Note that ~~,;;;J f(Xr) = ~iES f(i)V;(n). With Q a finite subset of S, and n; = J.Lj 1, the
unique stationary distribution,
::":
{"I
V;(n)
1
L.t :
iEQ
J.Lz
1 )} 11/lloo,
+ "(V;(n)
L.t   +:i<f.Q
J.Lz
where 11/lloo = sup{l/(i)l : i E S}. The sum over i E Q converges a.s. to 0 as n+ oo, by part (a).
The other sum satisfies
"(V;(n)
L.t  i<f.Q
n
1) _2 L.t
"(Vi(n)
iEQ
n
+
J.Li
+ lrj )
S.
33. (a) Since the chain is persistent, we may assume without loss of generality that Xo = j. Define
the times RJ, R2, ... of return to j, the sojourn lengths SJ, S2, ... in j, and the times VJ, V2, ...
between visits to j. By the Markov property and the strong law of large numbers,
1
n
"S
~
a.s.
1
1~
a.s.
 Rn =  L.t Vr + J.Lj.
n
n r=l
r~,
n r=l
gj
Also, Rn/ Rn+J ~ 1, since J.Lj = E(RJ) < oo. If Rn < t < Rn+l then
lot f(X(s))ds L l
=
l(X(s)=jJdS
jES 0
34. (a) By the first BorelCantelli lemma, Xn = Yn for all but finitely many values of n, almost
surely. Off an event of probability zero, the sequences are identical for all large n.
(b) This follows immediately from part (a), since Xn  Yn = 0 for all large n, almost surely.
(c) By the above, a;; 1 ~~~ (Xr  Yr) ~ 0, which implies the claim.
347
[7.11.35][7.11.37]
Solutions
by assumption (a), whence {Xn} and {Yn} are tailequivalent (see Problem (7.11.34)). By assumption
(b) and the martingale convergence theorem (7.8.1) applied to the partial sums 2:::~= 1 (Yn  IE(Yn)),
the infinite sum 2:::~ 1 (Yn  IE(Yn)) converges almost surely. Finally, I:~ 1 IE(Yn) converges by
assumption (c), and therefore 2:::~ 1 Yn, and hence 2:::~ 1 Xn, converges a.s.
lP'(/ns
= 1 for 1 .::5 s
.::5 r)
IT lP'Uns = 1).
s=1
Since F is continuous, the Xi take distinct values with probability 1, and furthermore the ranking of
X 1, X2, ... , Xn is equally likely to be any of the n! available. Let x1, x2, ... , Xn be distinct reals,
and write A= {Xi =Xi for 1 .::5 i .::5 n}. Now,
lP'Uns = 1 for 1 .::5 s .::5 r
I A)
1{(nns1
 1) (n 1 ns1)! }{(nsns21  1) (ns1 1 ns2)! } (n1 1)!
=I
n.
t(/~)
~0
1
J
 1
logn }= 1
The result follows on recalling that
as n+ oo.
oo.
37. By an application of the three series theorem of Problem (7.11.35), the series converges almost
surely.
348
Random processes
With ai (n)
+ n)
p(Xm,Xm+n)=r=r~~~~~~~~~==T=~~
Now, a1 (m)
+ n)(l 
a1 (m
+ n))
+ fJ (1
a+fJ
a  fJ) ,
.
1 n
hm  LlP'(Xr
1)
= .
a+ fJ
n+oo n r=l
The process is strictly stationary if and only if Xo has the stationary distribution.
2.
E(T(u)T(v))dudv=21 1 1v e 2}..(vu)dudv
~<u<v<t
v=O u=O
=2 {
=
3.
I1 ( t 
1
2A.
+ 21A. e 2At)
+ y) =
It
as t ~ oo.
+ y) > 0)
= lP'(N(x) > 0) + lP'({N(x) = 0} n
for x, y 2: 0.
:S g(x) + g(y)
lP'(N(x
349
{N(x
+ y) N(x)
> 0})
[8.3.1][8.3.4]
Solutions
Random processes
Such a function g is called subadditive, and the existence of A follows by the subadditive limit theorem
discussed in Problem (6.15.14). Note that A= oo is a possibility.
Next, we partition the interval (0, 1] into n equal subintervals, and let ln(r) be the indicator
function of the event that at least one arrival lies in ( (r  1) In, r In] , 1 .::5 r .::5 n. Then 2:::~= 1 In (r) t
N(l) as n + oo, with probability 1. By stationarity and monotone convergence,
E(N(l)) =
m:( lim
~ In(r))
n+oo L....t
lim
n+oo
r=I
m:(~
In(r))
L....t
r=I
2.
Bn+I = {
X  1 if Bn = 0,
Bn  1 if Bn > 0.
Therefore, B is a Markov chain with transition probabilities Pi,iI = 1 fori > 0, and POJ = fJ+I
for j 2: 0, where fn = lP'(X = n). The stationary distribution satisfies TCJ = TCJ+J + nof}+J, j 2: 0,
with solution n1 = lP'(X > j)IE(X), provided E(X) is finite.
The transition probabilities of B when reversed in equilibrium are
Pi,i+I
TCi+I
= 7; =
lP'(X > i + 1)
lP'(X > i)
fi+I
PiO = lP'(X > i),
fori 2: 0.
These are the transition probabilities of the chain U of Exercise (8.3.1) with the Jj as given.
3. We have that pnun = 2:::~=! pnkUnkPk fk, whence Vn = pnun defines a renewal sequence
provided p > 0 and l:n pn fn = 1. By Exercise (8.3.1 ), there exists a Markov chain U and a state s
such that Vn = lP'(Un = s)+ ns, as n+ oo, as required.
4.
oo
oo
oo
k=I
r=k
L E(N(r))sr = L L UkSr = L Uk L sr
r=O
r=I k=I
L....t1s
k=I
1s
1s
Let Sm = l:k=I Xk and So= 0. Then lP'(N(r) = n) = lP'(Sn .::5 r) lP'(Sn+I .::5 r), and
~sfE
00
[(N(t)k
00
= ~sf
00
Now,
350
00
n+
k _k
1)
lP'(Sn .::5 t) .
Solutions
Queues
[8.3.5][8.4.4]
L...J
[(N(t)k+k)] =
U(d
1s
5. This is an immediate consequence of the fact that the interarrival times of a Poisson process are
exponentially distributed, since this specifies the distribution of the process.
=2
'A
1t
1t }
1 { 1t
'A
'A }
'A + Jt . 'A + Jt + 'A + Jt + 2 'A + Jt . 'A + Jt + 'A + Jt
= 1
(1t)
= 2+
2'A~t
('A +
~t)2.
"A+~t
2'A~t
2. The given event occurs if the time X to the next arrival is less than t, and also less than the time
Y of service of the customer present. Now,
3.
and the result follows. If it takes a time b to cross the other lane, and so a + b to cross both, then, with
an obvious notation,
(a)
(b)
eal. 1
ebJ.L 1
IE(Ta) +IE(n) =   +   ,
'A
1t
e<a+b){A+J.L)  1
IE(Ta+b) =     "A+~t
The latter must be the greater, by a consideration of the problem, or by turgid calculation.
4.
'A(n + 1)
n + 2 Jl'n,
n:::: 0.
00
351
[8.4.5][8.5.5]
Solutions
Random processes
5. By considering possible transitions during the interval (t, t +h), the probability Pi (t) that exactly
i demonstrators are busy at time t satisfies:
P2(t +h)= P1 (t)2h
+ P2(t)(l
 2h)
+ o(h),
+ P1(t)h + o(h).
Hence,
P~ (t)
= 2pt (t)
 2p2(t),
Pi (t)
= 2po(t) 
Jp1 (t)
+ 2p2 (t),
p (t)
2po(t)
+ Pt (t),
and therefore P2(t) =a+ be 2t + ceSt for some constants a, b, c. By considering the values of P2
and its derivatives at t = 0, the boundary conditions are found to be a + b + c = 0,  2b  5c = 0,
4b + 25c = 4, and the claim follows.
Writing W(s)
..fiX, W(t)
If
+sin 1
t
= ,fiz,
If
and W(u)
+sin 1
u
If}
u
var(Z I X, Y) =
yielding var(W(t)
I W(s),
(ut)(ts)
t(u s)
(ut)W(s)+(ts)W(u)]
u_ s
W(u) W(s), W(v)
},
Whenever a 2 + b 2 = 1.
4.
Let lij(n)
= W((j +
5. They all have mean zero and variance t, but only (a) has independent normally distributed increments.
352
Solutions [8.7.1][8.7.3]
Problems
=0
2. We have, by iteration, that Yn = Sn (m) +am+ 1 Ynm1 where Sn (m) = LJ=O cxi Zn j. There
are various ways of showing that the sequence {Sn (m) : m ::: 1} converges in mean square and almost
surely, and the shortest is as follows. We have that cxm+ 1Ynm1 + 0 in m.s. and a.s. as m + oo;
to see this, use the facts that var(cxm+ 1 Ynm1) = cx 2(m+ 1) var(Yo), and
> 0.
It follows that Sn(m) = Yn am+ 1Ynm1 converges in m.s. and a.s. as m+ oo. A longer route to
the same conclusion is as follows. For r < s,
whence {Sn(m) : m ::: 1} is Cauchy convergent in mean square, and therefore converges in mean
square. In order to show the almost sure convergence of Sn(m), one may argue as follows. Certainly
whence Z:::f:=o cxi Zn j is a.s. absolutely convergent, and therefore a.s. convergent also. We may
express limm+oo Sn (m) as Z:::f:=o cxi Zn j. Also, am+ 1Ynm1 + 0 in mean square and a.s. as
m + oo, and we may therefore express Yn as
00
Yn =
L cxi Zn
a.s.
j=O
It follows that E(Yn) = limm+oo E(Sn(m)) = 0. Finally, for r > 0, the autocovariance function
is given by
c(r) = cov(Yn, Ynr) = E{ (cxYn1 + Zn)Ynr} = cxc(r  1),
whence
cxlrl
c(r)
= cxlrlc(O) =  2 ,
lex
1ft is a nonnegative integer, N(t) is the number ofO's and 1's preceding the (t+ l)th 1. Therefore
k:::t+l.
353
[8.7.4][8.7.6]
4.
Solutions
Random processes
We have that
A.h
= i) = {
+ o(h)
+ o(h)
+ 1,
= i  1,
if j = i
if j
11ih
1 (A.+ 11i)h
+ o(h)
if j = i,
an immigrationdeath process with constant birth rate A. and death rates /1i = i fl.
Either calculate the stationary distribution in the usual way, or use the fact that birthdeath
processes are reversible in equilibrium. Hence 'Ani = 11U + l)rri+l fori 2:: 0, whence
nl = .,1
I.
(A.)i
eA/JL
11
i :::: 0.
'
5. We have that X(t) = R cos(IJ!) cos(&t)  R sin(IJ!) sin(&t). Consider the transformation u
r cos 1/f, v = r sin 1/f, which maps [0, oo) x [0, 2n) to R 2 . The Jacobian is
au
au
ar
a!fr
ar
aljf
av
av
= r,
whence U = R cos IJf, V =  R sin IJf have joint density function satisfying
rfu,v(rcosljf, rsin 1/f)
Substitute fu,v(u, v)
= fR,\jl(r, 1/f).
Thus Rand IJf are independent, the latter being uniform on [0, 2rr).
6. A customer arriving at time u is designated green if he is in state A at time t, an event having
probability p(u, t  u). By the colouring theorem (6.13.14), the arrival times of green customers form
a nonhomogeneous Poisson process with intensity function A.(u)p(u, t  u), and the claim follows.
354
Stationary processes
Wn
LakZnk
k=O
for the real sequence {ak : k 2:: 0}. Substitute, to obtain ao = l, a1 =a, ar = aar1
with solution
+ fJar2 r
2:: 2,
otherwise,
where A. 1 and A.2 are the (possibly complex) roots of the quadratic x 2 ax  fJ = 0 (these roots are
distinct if and only if a 2 + 4{3 =!= 0).
Using the method in the solution to Problem (8.7.2), the sum in(*) converges in mean square and
almost surely if IA.1I < 1 and IA.2I < 1. Assuming this holds, we have from(*) that JE(Wn) = 0 and
the autocovariance function is
c(m) = JE(Wn Wnm) = ac(m 1)
+ {Jc(m 2),
::=::
1,
by the independence of the Zn. Therefore W is weakly stationary, and the autocovariance function
may be expressed in terms of a and fJ.
2.
We adopt the convention that, if the binary expansion of U is nonunique, then we take the
(unique) nonterminating such expansion. It is clear that Xi takes values in {0, 1}, and
for all x1, x2, ... , Xn; therefore the X's are independent Bernoulli random variables. For any sequence k1 < k2 < ... < kr, the joint distribution of vk,' vk2' ... ' vkr depends only on that of
Xk 1+1, Xk 1+2 .... Since this distribution is the same as the distribution of X 1, X2, ... , we have that
(Vk 1 , Vk2 , ... , Vkr) has the same distribution as (Vo, Vk2k1 , . , Vkrk! ). Therefore V is strongly
stationary.
Clearly JE(Vn) = JE(Vo) = i. and, by the independence oftlie Xi,
00
cov(Vo, Vn)
355
tz(i)n.
[9.1.3][9.2.1]
Solutions
Stationary processes
3. (i) For meansquare convergence, we show that Sk = L~=O anXn is meansquare Cauchy convergent as k + oo. We have that, for r < s,
lE{ (Ss  Sr ) 2 }
= . _L
s
L
s
}2
since lc(m) I .::; c(O) for all m, by the CauchySchwarz inequality. The last sum tends to 0 as r, s + oo
if Li lai I < oo. Hence Sk converges in mean square ask + oo.
Secondly,
lE(
lakXkl) .::::
k=I
k=I
lakl
k=I
which converges as n + oo if the lakl are summable. It follows that Lk=I lakXkl converges
absolutely (almost surely), and hence Lk=I akXk converges a.s.
(ii) Each sum converges a.s. and in mean square, by part (i). Now
00
cy(m) =
ajakc(m + k j)
j,k=O
whence
L lcy(m)l.:::: c(O){ f
j=O
4. Clearly Xn has distribution :rr for all n, so that {f(Xn) : n :::: m} has fdds which do not depend
on the value of m. Therefore the sequence is strongly stationary.
f3
c(O)c(2)  c(1) 2
y = c(0)2  c(1)2 ;
356
Solutions
= c(2) = 0.
Therefore Xn+1
[9.2.2][9.3.2]
= Xn+1
= 0, and D = 0.
In both (a) and (b), little of substance is gained by using Xn+1 in place of Xn+1
2. Let {Zn : n = ... , 1, 0, 1, ... } be independent random variables with zero means and unit
variances, and define the movingaverage process
Zn + aZn1
~.
Xn=
By the projection theorem, Xn  Xn is orthogonal to the collection {Xnr : r > 1}, so that
00
.
E{(Xn Xn)Xnr} = 0, r :=:: 1. Set Xn = l::s=1 hsXns to obtam that
for s:::: 2,
where a = aj(l + a 2 ). The unique bounded solution to the above difference equation is hs
(l)s+ 1as, and therefore
00
Xn
cov(Xn, Xnm)
brbsc(m
+ r  s),
m ::::0,
r,s=1
so that
X is weakly stationary.
9.3 Solutions. Autocovariances and spectra
1.
cos(nA.),
2.
= (eit:rr
cov(Xm, Xm+n)
= E(XmXm+n) =
E(ei{UVmU+V(m+n)J)
357
einJ... dF(A.),
Also
= 1/lv(n),
[9.3.3][9.3.4]
Solutions
Stationary processes
3.
(i)
p(t)
(ii)
p(t) =
4.
1 2
= ezt
c~it
+ 1 ~it)
= 1
~ t2 .
var ( 1 n Xj ) = 21
n }=!
Ln
n j,k=1
1 (L
n
) dF(A.).
ez(jk))..
j,k=1
(Jr,Jr]
The integrand is
1  cos(nA.)
1 cos A. '
whence
~X )
var ( 1 L
n }= 1 1
It is easily seen that I sine I .:5
= c(O)
1 (
(Jr,Jr]
sin(nA./2) ) 2 dF(A.).
n sin(A./2)
A./2 ) 2
1 2
sin(A./2)
.:5 (2n)
As n + oo, the integrand converges to the function which is zero everywhere except at the origin,
where (by continuity) we may assign it the value 1. It may be seen, using the dominated convergence
theorem, that the integral converges to F (0)  F (0), the size of the discontinuity of F at the origin,
and therefore the variance tends to 0 if and only if F(O)  F(O) = 0.
Using a similar argument,
1 n1
L c(j) =
n }=0
(0)
(n1
)
_ceij).. dF(A.) = c(O)
gn(A.)dF(A.)
n
(Jr,Jr] j=O
(Jr,Jr]
where
ifA.
= 0,
ifA.
= 0,
gn(A.) = { 1ein).. 1
n(ei)..  1)
is a bounded sequence of functions which converges as before to the Kronecker delta function 8;..o.
Therefore
1 n1
c(j) + c(O) ( F(O)  F(O))
as n+ oo.
n }=0
L
358
Solutions
[9.4.1][9.5.2]
_ 1
lE.(YmY n) =
(n,n]
eimJ...einJ...
/(A.) dA. = 8mn
2rrf(A.)
This proves that such a sequence Xn may be expressed as a moving average of an orthonormal
sequence.
3. Let H x be the space of all linear combinations of the Xn, together with all limits of (meansquare) Cauchyconvergent sequences of such combinations. Using the result of Problem (7 .11.19),
all elements in H x are normally distributed. In particular, all increments of the spectral process are
normal. Similarly, all pairs in H x are jointly normally distributed, and therefore two members of
H x are independent if and only if they are uncorrelated. Increments of the spectral process have
zero means (by Exercise (9.4.1)) and are orthogonal. Therefore they are uncorrelated, and hence
independent.
1,
so that U'f A; E 1.
2. The lefthand side is the sum of covariances, c(O) appearing n times, and c(i) appearing 2(n i)
times for 0 < i < n, in agreement with the righthand side.
Let E > 0. Ifc(j) =
when j 2::: J. Now
1 2:{~~
2n
2{1
00,
j=1
j(a2
+E) }
+ a2
+E
j=l+1
as n + oo. A related lower bound is proved similarly, and the claim follows since E ( > 0) is arbitrary.
359
[9.5.3][9.6.3]
3.
Solutions
Stationary processes
It is easily seen that Sm = l::~o ai Xn+i constitutes a martingale with respect to the X's, and
oo
E(S~)
I>tE(X~+i):::;
L:>f,
i=O
i=O
Now n 1cm + 0 as n + oo, so that Z is defined in terms of the subsequence Xm+I, Xm+2, ...
for all m, which is to say that Z is a tail function of a sequence of independent random variables.
Therefore Z is a.s. constant, and so Z = 0 a.s.
where
to= 0.
2. For s, t :::: 0, X (s) and X (s + t) have a bivariate normal distribution with zero means, unit
variances, and covariance c(t). It is standard (see Problem (4.14.13)) that E(X(s + t) I X(s)) =
c(t)X(s). Now
c(s
by the Markov property. Therefore c satisfies c(s + t) = c(s)c(t), c(O) = 1, whence c(s) = c(l)lsl =
plsl. Using the inversion formula, the spectral density function is
Note that X has the same autocovariance function as a certain autoregressive process. Indeed,
stationary Gaussian Markov processes have such a representation.
3. If X is Gaussian and strongly stationary, then it is weakly stationary since it has a finite variance.
Conversely suppose X is Gaussian and weakly stationary. Then c(s, t) = cov(X(s), X(t)) depends
360
Solutions
Problems
[9.6.4][9.7.2]
on t  s only. The joint distribution of X(t1), X(t2), ... , X(tn) depends only on the common mean
and the covariances c(ti, fj ). Now c(ti, fj) depends on fj  ti only, whence X (tl), X (t2), ... , X Un)
have the same joint distribution as X(s + t1), X(s + t2), ... , X(s + tn). Therefore X is strongly
stationary.
4.
whence
cov(X(s) 2 , X(s
+ t) 2 ) =
=
+ t) 2)  1
m:{ X(s) 2 E(X(s + t) 2 I X(s))} 1
c(t) 2E(X(s) 4 ) + (1 c(t) 2 )E(X(s) 2 ) 1 =
E(X(s) 2 X(s
by an elementary calculation.
(b) Likewise cov(X(s) 3 , X(s + t) 3) = 3(3
2c(t) 2
+ 2c(t) 2 )c(t).
if k = 0,
_1_+_a_2__
=2'f3_2
1/3
{
c(k) =
f31kl1
2 }
a( 1 + af3 f3 )
1 f32
if k
0.
Set Yn+ 1 = 2:::~0 ai Yni and find the ai for which it is the case that E{ (Yn+ 1  Yn+ 1) Ynk} = 0
for k ::::: 0. These equations yield
00
+ 1) =
c(k
L aic(k i),
k::::: 0,
i=O
::::: 0.
aipx(n) = ap
ajakpy(n
+ k j).
j,k=O
Therefore
2
aifx(A.)=
~;
oo
einJ...
n=oo
2
ajakpy(n+kj)
j,k=O
oo
= ay ""'aakei(kj)J...
2n L
j,k=O
""' ei(n+kj)J...py(n+kj)
n=oo
2
iJ... 2
= ayiGa(e
)I fy(A.).
361
[9.7.3][9.7.5]
Solutions
Stationary processes
IAI
< n,
where c = a{; I a
3.
Xn
= Yn
 Yn
= Yn aYn1
 f3Yn2
Now Xn is orthogonal to {Ynk : k 2: 1}, so that the Xn are uncorrelated random variables with
spectral density function fx(A) = (2n) 1 , A E (n, n). By the result of Problem (9.7.2),
al fx (A) = a{; 11 
whence
f y (A)
a2ja2
X
n <A< n.
4. Let {X~ : n 2: 1} be the interarrival times of such a process counted from a time at which a
meteorite falls. Then Xi, X~, ... are independent and distributed as X2. Let Y~ be the indicator
function of the event {X:n = n for some m}. Then
E(YmYm+n)
= lP'(Ym = 1, Ym+n = 1)
= lP'(Ym+n = 1 I Ym = l)lP'(Ym =
1)
= lP'(Y~ =
l)a
a{lP'(Y~
, c(n) = Re {
1
~
,
1 } .
fy(A) =  1 ~
L..J eznA
L..J eznAc(n)2n n=oo
a(l a)
na(l a) n=O
2n
Now
00
00
n=O
n=O
L einl. y~ = L eiAT~
where T~
when eil.
i=
5.
Therefore
1
n(la)
Re {
1
1(jJ(A)
a . }  1,
 1e11.
2n
We have that
E(cos(nU)) =
:n:
:n: 2n
cos(nu) du = 0,
362
IAI
< n.
1) a},
Solutions
Problems
[9.7.6][9.7.7]
n)U1)} = 0
if m =f. n. Hence X is stationary with autocorrelation function p(k) = 8ko, and spectral density
function f(J....) = (2n) 1 for IJ....I < n. Finally
E{ cos(mU) cos(nU) cos(rU)}
= i{p(m
+ n r) + p(m n r)}
which takes different values in the two cases (m, n, r) = (1, 2, 3), (2, 3, 4).
6. (a) The increments of N during any collection of intervals {(u;, v;) : 1 :::0 i :::: n} have the same
fdds if all the intervals are shifted by the same constant. Therefore X is strongly stationary. Certainly
E(X(t)) = J....a for all t, and the autocovariance function is
c(t) = cov(X(O), X(t)) = { 0
J....(a t)
if t >a,
ifO :::0 t :::0 a.
p(t)
0
1 itfai
if it I >a,
if it I Sa,
2(J cov(X(s), X(t)) = 2(J cov(W(s) W(s 1), W(t) W(t 1))
(s 1)
ift1<sst.
This depends on t  s only, and therefore X is stationary; X is Gaussian and therefore strongly
stationary also.
The autocorrelation function is
p(h) =
0
1 lhl
if lhl :::: 1,
if lhl < 1,
= (1
cos J....)j(nJ.... 2 ).
7. We have from Problem (8.7.1) that the general movingaverage process of part (b) is stationary
with autocovariance function c(k) = '}=o a}ak+ J, k :::: 0, with the convention that as = 0 if s < 0
ors > r.
(a) In this case, the autocorrelation function is
p(k)={ 1 +a 2
0
363
if k = 0,
if lkl = 1,
iflk1>1,
[9.7.8][9.7.13]
Solutions
Stationary processes
IA.I < n.
/(A.)=__!__
00
eik)..p(k)
2n k=oo
8.
')..
IA(ez )I
2nc(O)
where c(O) =
f(x)
= 1
100 eztx
. p(t) dt
2n _ 00
2n oo
and
__!__
100
2n oo
The integrand is dominated by the integrable function 21p(t)l. Using the dominated convergence
theorem, we deduce that lf(x +h) f(x)l+ 0 ash+ 0, uniformly in x.
By Exercise (9.5.2), var(n 1 'L,'J=l XJ) + a 2 if Cn = n 1 'Lj:J cov(Xo, Xj) + a 2 . If
cov(Xo, Xn) + 0 then Cn + 0, and the result follows.
9.
10. Let X 1, X 2, ... be independent identically distributed random variables with mean p,. The sequence X is stationary, and it is a consequence of the ergodic theorem that n 1 'L,'J=I 1 + Z a.s.
and in mean, where Z is a tail function of X 1, X 2 ... with mean p,. Using the zeroone law, Z is a.s.
constant, and therefore lP'(Z = p,) = 1.
11. We have from the ergodic theorem that n 1 'Lt=l Yi + lE(Y I 1) a.s. and in mean, where 1 is
the afield of invariant events. The condition of the question is therefore
lE(Y 11) = lE(Y)
Suppose(*) holds. Pick A E 1, and set Y = lA to obtain lA = Q(A) a.s. Now lA takes the values 0
and 1, so that Q(A) equals 0 or 1, implying that Q is ergodic. Conversely, suppose Q is ergodic. Then
lE(Y I 1) is measurable on a trivial afield, and therefore equals lE(Y) a.s.
12. Suppose Q is strongly mixing. If A is an invariant event then A = rnA. Therefore Q(A) =
Q(A n rnA) + Q(A) 2 as n + oo, implying that Q(A) equals 0 or 1, and therefore Q is ergodic.
13. The vector X = (X 1, X2, ... ) induces a probability measure Q on (l~T, JF,T). Since Tis measurepreserving, Q is stationary. Let Y : rntT + rnt be given by Y(x) = x1 for x = (xl, x2, .. .), and define
Yi (x) = Y(ril (x)) where r is the usual shift operator on rntT. The vector Y = (f1, Y2, ... ) has the
same distributions as the vector X. By the ergodic theorem for Y, n 1 'Lt=l Yi + lE(Y I 'J.) a.s. and
in mean, where '!f. is the invariant a field of r. It follows that the limit
1 n
Z = lim  ""Xi
n+oo n L...t
i=l
364
Solutions
Problems
implying that U(w) = U(Tw) a.s. It follows that U is 1measurable, and it is the case that Z = U
a.s. Take conditional expectations of(*), given 1, to obtain U = E(X 11) a.s.
If Tis ergodic, then 1 is trivial, so that E(X 11) is a.s. constant; therefore E(X 11) = E(X) a.s.
This proves that A is independent of both [0, 1) and [1, 1). A similar proof gives that A is independent
of any set E which is, for some n, the union of intervals of the form [k2n, (k+ 1)2n) forO ::::: k < 2n.
It is a fundamental result of measure theory that there exists a sequence E 1, Ez, ... of events such that
(a) En is of the above form, for each n,
(b) lP'(A LEn) + 0 as n + oo.
Choosing the En accordingly, it follows that
lP'(A n En) = lP'(A)lP'(En) + lP'(A) 2
by independence,
15. We may as well assume that 0 < a < 1. LetT : [0, 1) + [0, 1) be given by T(x) = x +a
(mod 1). It is easily seen that Tis invertible and measurepreserving. Furthermore T(X) is uniform
on [0, 1], and it follows that the sequence Z 1, Zz, ... has the same fdds as Zz, Z3, ... , which is to
say that Z is stationary. It therefore suffices to prove that T is an ergodic shift, since this will imply
by the ergodic theorem that
1 n
{I
Zj + E(ZI) =
g(u) du.
n j=l
o
L
Jo
We use Fourier analysis. Let A be an invariant subset of [0, 1). The indicator function of A has
a Fourier series:
00
IA(x) ~
anen(x)
n=oo
1
an=
2n
lolo
IA(x)en(x)dx = 1

2n
365
en(x)dx.
[9.7.16][9.7.18]
Solutions
Stationary processes
= T(x),
Now I A = I r1 A since A is invariant. We compare the previous formula with that of (* ), and deduce
that an = e 2 nina an for all n. Since a is irrational, it follows that an = 0 if n =f 0, and therefore I A
has Fourier series ao, a constant. Therefore lA is a.s. constant, which is to say that either lP'(A) = 0
or lP'(A) = 1.
16. Let G 1 (z) = E(zX(t)), the probability generating function of X(t). Since X has stationary
independent increments, for any n (~ 1), X(t) may be expressed as the sum
n
= eJ..(t)(lA(z))
= {X(s) X(O)}
+ { X(t) X(s) },
0:::::
s::::: t,
whence, by stationarity,
{m(t) m(O)} = {m(s) m(O)} + {m(t s) m(O)}.
Now m is continuous, so that m(t) m(O) = {Jt, t ~ 0, for some {3; see Problem (4.14.5).
(b) Take variances of(*) to obtain v(t) = v(s) + v(t s), 0 ::::: s ::::: t, whence v(t) = a 2 t for some
(J2.
18. In the context of this chapter, a process Z is a standard Wiener process if it is Gaussian with
Z(O) = 0, with zero means, and autocovariance function c(s, t) = min{s, t}.
366
Solutions [9.7.19][9.7.20]
Problems
(a) Z(t)
cov(Z(s), Z(t))
=(s+a)aa+a=s,
s~t.
= 0, JE(Z(t)) = 0.
19. The process W has stationary independent increments, and G(t) = JE(IW(t)l 2 ) satisfies G(t)
t + 0 as t + 0; hence
(u) d W (u) is well defined for any satisfying
jgo
Sn(t)
n~l.
}=0
However Sn(t) = W(t) for all n, and therefore Sn(t) ~ W(t) as n+ oo.
Finally, Y (s) is the limit (in meansquare) of a sequence of normal random variables with mean
0, and therefore is Gaussian with mean 0. If s < t,
cov(Y(s), Y(t)) = fooo (e(su) l[O,sJ(u)) (e(tu) I[o,t](u)) dG(u)
los
e 2ust d u_ 21 ( e st e st) .
0
Y is an OmsteinUhlenbeck process.
20. (a) W(t) is N(O, t), so that
JEIW(t)l =
oo
00
var(IW(t)l)
lui
ez<u ft) du =
..;z;rt
2)
= JE(W(t) 2 )  2t
n = t ( 1;
367
J2ili(,
.
Solutions
Stationary processes
The process X is never negative, and therefore it is not Gaussian. It is Markov since, if s < t and
B is an event defined in terms of {X(u) : u ::5 s}, then the conditional distribution function of X(t)
satisfies
1 Jiiit
00
eU
ez(u ft) du
= e2 1
00
Secondly, W(s) + W(t) = 2W(s) + {W(t) W(s)} is N(O, 3s + t) if s < t, implying that
E(Y(s)Y(t)) = JE: (eW(s)+W(t)) = A(3s+t),
and therefore
s < t.
W(1) is N(O, 1), and therefore Y(1) has the lognormal distribution. Therefore Y is not Gaussian.
It is Markov since W is Markov, and Y(t) is a one{Jne function of W (t).
(c) We shall assume that the random function W is a.s. continuous, a point to which we return in
Chapter 13. Certainly,
E(Z(t)) =lot E(W(u)) du
E(Z(s)Z(t))
!1
= 0,
E(W(u)W(v))dudv
O<u<s
o;v;t
=1s
u=O
{fu
vdv+ t
udv}du=is (3ts),
lv=O
lv=u
2
s<t,
n+oo
~ (!_)
W(itjn),
n
L...J
i=l
each such summation being normal. The limit of normal variables is normal (see Problem (7.11.19)),
and therefore Z(t) is normal. The limit in(*) exists a.s., and hence in probability. By an appeal to
(7.11.19b), pairs (Z(s), Z(t)) are bivariate normal, and a similar argument is valid for all ntuples of
the Z(u).
The process Z is not Markov. An increment Z (t)  Z (s) depends very much on W (s) = Z' (s ),
and the collection {Z (u) : u ::; s} contains much information about Z' (s) in excess of the information
contained in the single value Z(s).
21. Let Ui
X(ti). The random variables A= U1, B
Uz U1o C
U3 Uz, D
U4 U3
are independent and normal with zero means and respective variances t1, tz  t1, t3  tz, t4  t3. The
Jacobian of the transformation is 1, and it follows that U1, Uz, U3, U4 have joint density function
e!Q
/u(u) = (2n) 2,Jtl(t2 t1)(t3 tz)(t4 t3)
368
Solutions
Problems
where
ui
(u2 u1) 2
fl
f2  tl
Q=+
Likewise ul and
(u3  u2) 2
t3  t2
(u4 u3) 2
+'~
f4  t3
where
)2
2
2
(
S=~+ U3U2 +~.
f2  fl
t3  t2
f4  f3
Now g is the density function of a bivariate normal distribution with zero means, marginal variances
(t2  tl )(t3  t2)
t3 tl
and correlation
1
x(lx)
var(Fn(x)) = var(h(x)) =
.
n
n
By the central limit theorem, ,Jn{Fn(x) x} ~ Y(x), where Y(x) is N(O, x(1 x)).
(b) The limit distribution is multivariate normal. There are general methods for showing this, and
here is a sketch. If 0 ::::= x1 < x2 ::::= 1, then the number M2 (= nF(x2)) of the Ij not greater than
x2 is approximately N(nx2, nx2(l x2)). Conditional on {M2 = m}, the number M1 = nF(xl)
is approximately N(mu, mu(l  u)) where u = xlfx2. It is now a small exercise to see that the
pair (M1, M2) is approximately bivariate normal with means nx1, nx2, with variances nx1 (1  x1),
nx2(l  x2), and such that
whence cov(Ml, M2) ~ nx1 (1  x2). It follows similarly that the limit of the general collection is
multivariate normal with mean 0, variances x;(1 x;), and covariances Cij = x;(l Xj).
(c) The autocovariance function of the limit distribution is c(s, t) = min{s, t}  st, whereas, for
0:::: s :::: t :::: 1, we have that cov(Z(s), Z(t)) = s ts st + st = min{s, t} st. It may be shown
that the limit of the process { Jn (Fn (x)  x) : n :::: 1} exists as n + oo, in a certain sense, the limit
being a Brownian bridge; such a limit theorem for processes is called a 'functional limit theorem'.
369
10
Renewals
SincelE(Xl) > 0, there exists E (> 0) such thatlP'(X1 2: E)> E. LetXIc = El{xk::::E} and denote
by N' the related renewal process. Now N(t) :s N'(t), so that lE(eeN(t)) :s lE(eeN'{tl), fore > 0.
Let Zm be the number of renewals (in N') between the times at which N' reaches the values (m  1)E
and mE. The Z's are independent with
Eee
JE(ee Zm) _ ~
 1  (1  E)ee'
whence JE(eeN'(t)) :S (Eee {1 (1 E)ee}  1)t/E for sufficiently small positive e.
2. Let X 1 be the time of the first arrival. If X 1 > s, then W = s. On the other hand if X 1 < s, then
the process starts off afresh at the new starting time X 1. Therefore, by conditioning on the value of
x1,
Fw(x) =
fooo lP'(W :s
=los
I x1
=los
= u)dF(u)
lP'(W :S x u)dF(u)
lP'(W
:s X  u)dF(u) +
00
1 dF(u)
+ {1 F(s)}
if x 2: s. It is clear that F w (x) = 0 if x < s. This integral equation for F w may be written in the
standard form
Fw(x) = H(x)
where H and
F are given by
H(x) = {
~ F(s)
if X <
S,
ifx:=::s,
F(x) =
{ F(x)
F(s)
if x < s,
if X 2: s.
This renewaltype equation may be solved in the usual way by the method ofLaplaceStieltjes transforms. We have that Fw = H* + FiVF*, whence Fw = H* j(l F*). If N is a Poisson process
then F(x) = 1 el.x. In this case
H*(e) =
370
Solutions
Limit theorems
so that
(A.+ e)e(A+O)s
Fw(e)
[10.1.3][10.2.2]
+ A.e(A+O)s
We have as usual that lP'(N(t) = n) = lP'(Sn :S t) lP'(Sn+1 :S t). In the respective cases,
(a)
  Jor {;,.nbxnb1
(b)
4.
lP'(N(t) n)
r(nb)
;,.(n+1)bx(n+1)b1} A.x
f((n + l)b)
e
dx.
O:::;t:::;l.
Hence m' = 1 + m, with solution m(t) = e1  1, for 0 ::::; t ::::; 1. (For larger values oft, m(t) =
1 + m(t x) dx, and a tiresome iteration is in principle possible.)
JJ
O:::;t:::;l.
1.
+ oo a.s.
We have that
since {M :S i  1} is defined in terms of Z 1, Z2, ... , Zi 1, and is therefore independent of Z;.
Similarly E(Z; ZJ l(M:;:jj) = E(ZJ )E(Zi l(M:;:jj) = 0 if i < j. It follows that
E(Tli) =
00
00
i=1
i=1
[10.2.3][10.2.5]
Solutions
Renewals
3. (i) The shortest way is to observe that N(t) + k is a stopping time if k :::: 1. Alternatively, we
have by Wald's equation that JE(TN(t)+I) = Jl(m(t) + 1). Also
k:::: 2,
jl,
lE(TN(t)+k) = lE(TN(t)+I)
+ 2:JE(XN(t)+j) =
Jl(m(t)
+ k).
j=2
(ii) Suppose p
=f: 1 and
li"(Xt=a)= {
Then f.l = 2  p
=f: 1.
if a= 1,
1 p
if a= 2.
Also
=f:
+ plE(Tt
N(1) = 1) = p,
Jlm(l).
4. Let V(t) = N(t) + 1, and let W1, W2, ... be defined inductively as follows. W1 = V(1), W2
is obtained similarly to Wt but relative to the renewal process starting at the V(1)th renewal, i.e., at
time TN(I)+1 and Wn is obtained similarly:
Wn = N(Txnl
+ 1) N(Txn_ 1 ) + 1,
n:::: 2,
t i=I
It follows that the family { m 1 2::~ 1 W; : m :::: 1} is uniformly integrable (see Theorem (7.10.3)).
Now N(t).:::; V(t), and so {N(t)/t: t:::: 0} is uniformly integrable also.
Since N (t) It ~ J.l 1, it follows by uniform integrability that there is also convergence in mean.
5.
(a) Using the fact that lP'(N(t) = k) = lP'(Sk .:::; t) li"(Sk+1 .:::; t), we find that
lE(sN(T)) =
lE(sN(T)) =
(v) (1  ~)
b
v+A.
v+A.
372
Excess life
Solutions
[10.3.1][10.3.3]
+ y) + g(y) dF(x).
Write h(x) = 1 F(x) to obtain g(y)h(t) = h(t + y), for y, t :=:: 0. With t = 0, we have that
g(y)h(O) = h(y), whence g(y) = h(y)/ h(O) satisfies g(t + y) = g(t)g(y), for y, t :=:: 0. Now g is
g(y)
= 1
F(t
Hence F(t)
1 eAt, and
2.
(a) Examine a sample path of E. If E(t) = x, then the sample path decreases (with slope 1)
until it reaches the value 0, at which point it jumps to a height X, where X is the next interarrival time.
Since X is independent of all previous interarrival times, the process is Markovian.
(b) In contrast, C has sample paths which increase (with slope 1) until a renewal occurs, at which they
drop to 0. If C (s) = x and, in addition, we know the entire history of the process up to time s, the
time of the next renewal depends only on the length ofthe spent period (i.e., x) of the interarrival time
in process. Hence C is Markovian.
3.
+ y)
l
G(t
+ y x)dm(x)
where G(u) = 1 F(u). Check the conditions of the key renewal theorem (10.2.7): g(t) = G(t + y)
satisfies:
(i) g(t) :::: 0,
(ii) f0 g(t) dt .:::= fcF[1  F(u)] du = E(X 1) < oo,
(iii) g is nonincreasing.
We conclude, by that theorem, that
00
1
lim lP'(E(t) _:::: y) = 1  
t+00
f1
100 g(x)dx =
0
loy [11
F(x)]dx.
0 f1
i oo
o
xr
1
[1F(x)]dx=11
11
laoo
xr+1
E(xr+1)
dF(x)=
1
.
M(r + 1)
o r+1
00
> t
+ y x)dm(x)dy,
t+oo
[10.3.4][10.3.5]
4.
Renewals
Solutions
We have that
lP'(E(t)>yjC(t)=x)=lP'(Xt>y+xiXt>x)=
1 F(y +x)
1 F(x)
whence
E(E(t)jC(t)=x)= foolF(y+x)dy=E{(Xtx)+}_
Jo
1 F(x)
1 F(x)
5. (a) Apply Exercise (10.2.2) to the sequence X; fL, 1 :::= i < oo, to obtain var(TM(t) tLM(t)) =
a 2 E(M(t)).
(b) Clearly TM(t) = t + E(t), where E is excess lifetime, and hence tLM(t) = (t + E(t)) (TM(t)tLM(t)), implying in tum that
fL 2 var(M(t)) =
where
SM(t)
SM(t)),
as t + oo
as t + oo.
This is valid under the weaker assumption that E(XI) < oo, as the following argument shows. By
Exercise (10.3.3c),
E(E(t) 2 ) = a(t) +lot a(t u) dm(u),
where a(u) = JE:({(Xt u)+} 2 ). Now use the key renewal theorem together with the fact that
a(t):::: E(XIl(x 1 >tJ)+ Oast+ oo.
Using the CauchySchwarz inequality,
as t + oo, by part (a) and (** ). Returning to (*), we have that
fL2
var(M(t))+
lim
t
t'>00
{a2
(m(t)+l) } =a2
.
t
{L
374
Solutions [10.4.1][10.5.3]
Renewalreward processes
= _A_e(A+JL)t + _1t_.
A+Jt
A+Jt
Hence the excess lifetime distribution is a mixture of the exponential distribution with parameter Jt,
and the distribution of the sum of two exponential random variables, thus,
where g(x) is the density function of a typical interarrival time. By Wald's equation,
E(t
We substitute
( + ~)
(m(t)
+ 1).
p 11 (t)
(A1+1t1) + (1 Pn(t))1t1=1t1 + A
E(E(t)) = Pn(t)
to obtain the required expression.
lot
1
t 0
a.s.
1
Jtjgj
l{X(s)=j)dS ~   = lfj.
where T1 (F) is the total time spent in F up to time t. Take the limit as t
F t S, to obtain the required result.
2. Suppose you are paid a reward at unit rate during every interarrival time of type X, i.e., at all
times tat which M(t) is even. By the renewalreward theorem (10.5.1),
1
lot I
M s is even ds ~
o { ()
3.
Suppose, at timet, you are paid a reward at rate C(t). The expected reward during an interval
(cycle) of length X is fox s ds = ~ X 2 , since the age C is the same at the time s into the interval. The
375
[10.5.4][10.6.3]
Solutions
Renewals
result follows by the renewalreward theorem (10.5.1) and equation (10.5.7). The same conclusion is
valid for the excess lifetime E(s), the integral in this case being fox (X s) ds = ~ X 2 .
4. Suppose Xo = j. Let V1 = min{n ::: 1 : Xn = j, Xm = k for some 1 :S m < n}, the first visit
to j subsequent to a visit to k, and let Vr+1 = min{n ::: Vr : Xn = j, Xm = k for some Vr + 1 :S
m < n}. The Vr are the times of a renewal process. Suppose a reward of one ecu is paid at every visit
to k. By the renewalreward theorem and equation (10.5.7),
The latter expectation in (*) is the mean of a random variable N having the geometric distribution
f'(N = n) = p(l  p)n 1 for n ::: 1, where p = f'(1) < Tk I Xo = k). Since E(N) = p 1, we
deduce as required that
1/f'(1) < Tk I Xo = k)
lfk =     "         
E(n
I Xo
= j) + E(1j
I Xo
= k)
00
k=1
a=O
m(t) =
00.
(c) It is easiest to use Exercise (10.1.1), which implies the stronger conclusion that the moment
generating function of N (t) is finite in a neighbourhood of the origin.
2.
,,
2
v(t)= Jo E{(N(tu)+1) }dF(u)= Jo {v(tu)+2m(tu)+1}dF(u).
Fix x
lit Then
376
= (A.t) 2 + A.t.
Solutions [10.6.4][10.6. 7]
Problems
wherea(t) = l<tfp,) +xv'ta2jp,3j. Now,
'===' +
a ..j(i(i)
as t+ oo,
X
~x)
__... <1>(x)
as t+ oo
as t+ oo
}y
1i
by Exercise (10.3.3a). The current and excess lifetimes have the same asymptotic distributions.
5. Using the lackofmemory property of the Poisson process, the current lifetime C(t) is independent
of the excess lifetime E(t), the latter being exponentially distributed with parameter J..... To derive the
density function of C(t) either solve (without difficulty in this case) the relevant integral equation, or
argue as follows. Looking backwards in time from t, the arrival process looks like a Poisson process
up to distance t (at the origin) where it stops. Therefore C(t) may be expressed as min{Z, t} where Z
is exponential with parameter J....; hence
J....eJ..s
!qt)(s) = { 0
C(t)
if s < t
f  ,
s > t,
+ E(t),
6. The ith interarrival time may be expressed in the form T + Zi where Zi is exponential with
parameter J..... In addition, Z1, Zz, ... are independent, by the lackofmemory property. Now
1 F(x)
= lP'(T + zl
> x)
= lP'(Zl
>X T)
= eA(xT)'
X~
T.
Taking into account the (conventional) dead period beginning at time 0, we have that
lP'(N(t)
~ k)
= lP'(kT
+ ~ Zi
~ k),
kT,
l=l
377
[10.6.8][10.6.10]
Solutions
Renewals
We have that
lP'(E(t) s y) = F(t
+ y)
lot {1
F(t
+ y x)}dm(x).
m(t
+ y) =
F(t
+ y) + Jo
whence, by subtraction,
t+y
{1F(t+yx)}dm(x).
lP'(E(t)sy)= t
It follows that
1P'(X 1 sx)=
{x
{x {1F(xy)}dm(y)dFL(l)
}z=O }y=l
= [FL(l) {x{1F(xy)}dm(y)]x
~0
r FL(l){1F(xl)}dm(l)
8. (a) Each interarrival time has the same distribution as the sum of two independent random variables
with the exponential distribution. Therefore N(t) has the same distribution as L~ M(t)J where M is a
Poisson process with intensity A. Therefore m(t) = ~JE(M(t)) ilP'(M(t) is odd). Now JE(M(t)) =
At, and
oo (At)Zn+leAt
lP'(M(t) is odd)="'"'
= ~eA.t(eA.t eAt).
~ (2n + 1)!
n=O
With more work, one may establish the probability generating function of N(t).
(b) Doing part (a) as above, one may see that iii(t) = m(t).
9.
Clearly C(t) and E(t) are independent if the process N is a Poisson process. Conversely, suppose
that C(t) and E(t) are independent, for each fixed choice oft. The event {C(t) 0:: y} n {E(t) 0:: x}
occurs if and only if E (t  y) ::=: x + y. Therefore
lP'(C(t) ::=: y)lP'(E(t) ::=: x) = lP'(E(t y) ::=: x
+ y).
Take the limit as t ~ oo, remembering Exercise (10.3.3) and Problem (10.6.4), to obtain that
G(y)G(x) = G(x + y) if x, y ::=: 0, where
00
G(u) =
Now 1 G is a distribution function, and hence has the lackofmemory property (Problem (4.14.5)),
implying that G(u) = eJcu for some A. This implies in tum that [1 F(u)]/ f.i = G'(u) = AeJcu,
whence f.i = 1/A and F(u) = 1 eJcu.
10. Clearly N is a renewal process if Nz is Poisson. Suppose that N is a renewal process, and write
A for the intensity of N 1 , and Fz for the interarrival time distribution of Nz. By considering the time
X 1 to the first arrival of N,
378
Solutions [10.6.11](10.6.12]
Problems
1
100 [1
F(u)] du = eAx
xf.L
100 [11
F2(u)]
du,
xf.L2
fx
00
11. (i) Taking transforms of the renewal equation in the usual way, we find that
m* e 
F*(e)
< )  1 F*(e)
where
as
1
1 F*(e)
+a 2 ) +o(e 2 )
and expand to obtain the given expression. A formal inversion yields the expression form.
(ii) The transform of the righthand side of the integral equation is
By Exercise (10.3.3), Ff:(e) = [1  F*(e)]/(J.Le), and(*) simplifies to m*(e)  (m*  m* F* F*)/(J.Le), which equals m*(e) since the quotient is 0 (by the renewal equation).
Using the key renewal theorem, as t ~ oo,
lo [1t
2JL
379
2JL
[10.6.13][10.6.16]
Solutions
Renewals
whence md* = Fd* + md* F*, the transform of the given integral equation.
(ii)ArguingasinProblem(10.6.2), vd* = Fd*+2m*Fd*+v*Fd*wherev* = F*(l+2m*)/(lF*)
is the corresponding object in the ordinary renewal process. We eliminate v* to find that
13. Taking into account the structure of the process, it suffices to deal with the case I = 1. Refer
to Example (10.4.22) for the basic notation and analysis. It is easily seen that fJ = (v  1)A.. Now
F(t) = 1 evA.t. Solve the renewal equation (10.4.24) to obtain
+lot
g(t) = h(t)
h(t x) diii(x)
where iii(x) = vA.x is the renewal function associated with the interarrival time distribution
Therefore g(t) = 1, andm(t) = ef3t.
F.
14. We have from Lemma (10.4.5) that p* = 1  Fz + p* F*, where F* = Fy Fz. Solve to obtain
*
p
1 Fz
1 F*F*
y z
15. The first locked period begins at the time of arrival of the first particle. Since all future events
may be timed relative to this arrival time, we may take this time to be 0. We shall therefore assume
that a particle arrives at 0; call this the Oth particle, with locking time Yo.
We shall condition on the time X 1 of the arrival of the next particle. Now
lP'(L > t
I X1
= u) = {
lP'(Yo>t)
ifu>t,
ifu ~ t,
where L' has the same distribution as L; the second part is a consequence of the fact that the process
'restarts' at each arrival. Therefore
lP'(L > t) = (1 G(t))lP'(X1 > t)
+lot
16. (a) It is clear that M(tp) is a renewal process whose interarrival times are distributed as X1 +
Xz + + XR where lP'(R = r) = pqr 1 for r:::: 1. It follows that M(t) is a renewal process whose
first interarrival time
X(p) = inf{t : M(t) = 1} = p inf{t : M(tp) = 1}
380
Solutions [10.6.17][10.6.19]
Problems
r=1
oo
pqr14J(pt{
pqy(pt)
1  qqy(pt)
r=1
where qy is the characteristic function of F. Now qy (pt) = 1 + i Jlpt + o(p) as p ,), 0, so that
QJ (t)
1 + ijlpt + o(p)
1itLt+o(1)
1 + o(l)
1itLt
as p ,), 0. The limit is the characteristic function of the exponential distribution with mean fL, and the
continuity theorem tells us that the process M converges in distribution as p ,), 0 to a Poisson process
with intensity 1I fL (in the sense that the interarrival time distribution converges to the appropriate
limit).
(c) If M and N have the same fdds, then c/Jp(t) = qy(t), which implies that qy(pt) = qy(t)f(p + qqy(t)).
Hence 1/f(t) = c/J(t) 1 satisfies 1/f(pt) = q + pl/f(t) fort E JR. Now 1/f is continuous, and it follows as
in the solution toProblem(5.12.15) that 1/f has the form 1/f(t) = 1+,Bt, implyingthatqy(t) = (1 +.Bt) 1
for some .B E :::. The only characteristic function of this form is that of an exponential distribution,
and the claim follows.
17. (a) Let N(t) be the number of times the sequence has been typed up to the tth keystroke. Then N is
a renewal process whose interarrival times have the required mean fL; we have that JE(N(t))/t + fL 1
as t + oo. Now each epoch of time marks the completion of such a sequence with probability
( 1 0 ) 14 , so that
~lE(N(t)) = ~
t
t (1) (1)
14+
n= 14
100
100
14
as t+ oo,
19. We use the renewalreward theorem. The mean time between expeditions is B fL, and this is
the mean length of a cycle of the process. The mean cost of keeping the bears during a cycle is
B(B  1)cfL, whence the longrun average cost is {d + B(B  1)ctL/2}/(B fL).
381
11
Queues
1f P
lfn+1
+1+p
for n > 2,

with l::~o :rri = 1, have the given solution. If p ;::: 1, no such solution exists. It is slightly shorter to
use the fact that such a walk is reversible in equilibrium, from which it follows that 1f satisfies
for n ;::: 1.
2. (i) This continuoustime walk is a Markov chain with generator given by g01 = Bo, gn,n+ 1 =
Bnp/(1 + p) and gn,n1 = Bn/(1 + p) for n ;::: 1, other offdiagonal terms being 0. Such a process
is reversible in equilibrium (see Problem (6.15.16)), and its stationary distribution v must satisfy
Vngn,n+1 = Vn+1gn+1,n These equations may be written as
for n ;::: 1.
These are identical to the equations labelled(*) in the previous solution, with :rrn replaced by vnBn. It
follows that Vn = C:rrn/Bn for some positive constant C.
(ii) If Bo = A., Bn =A.+ f.i for n ;::: 1, we have that
I Q)}
=E{ (f.t~it)Q}
c_1_,...,P, = (1 _ p)
1pf.t/({tit)
382
+p
1i  A. ) .
{tA.it
Solutions
M/M/1
[11.2.4][11.2.7]
This is the characteristic function of the given distribution. The atom at 0 corresponds to the possibility
that Q = 0.
4. We prove this by induction on the value of i + j. If i + j = 0 then i = j = 0, and it is easy to
check that rr(O; 0, 0) = 1 and A(O; 0, 0) = 1, A(n; 0, 0) = 0 for n ::: 1. Suppose then that K ::: 1,
and that the claim is valid for all pairs (i, j) satisfying i + j = K. Let i and j satisfy i + j = K + 1.
The last ball picked has probability i /(i + j) of being red; conditioning on the colour of the last ball,
we have that
rr(n;i,j)= "+i .rr(n1;i1,j)+ "+j .rr(n+1;i,j1).
l
1
l
1
Now (i  1) + j = K = i + ( j  1). Applying the induction hypothesis, we find that
rr(n; i, j) = i
~j
+ i
~j
Substitute to obtain the required answer, after a little cancellation and collection of terms. Can you
see a more natural way?
5.
Let A and B be independent Poisson process with intensities J.... and p, respectively. These processes
generate a queueprocess as follows. At each arrival time of A, a customer arrives in the shop. At
each arrivaltime of B, the customer being served completes his service and leaves; if the queue
is empty at this moment, then nothing happens. It is not difficult to see that this queueprocess is
M(J....)/M(p,)/1. Suppose that A(t) = i and B(t) = j. During the timeinterval [0, t], the order of
arrivals and departures follows the schedule of Exercise (11.2.4), arrivals being marked as red balls
and departures as lemon balls. The imbedded chain has the same distributions as the random walk
of that exercise, and it follows that JP>(Q(t) = n I A(t) = i, B(t) = j) = rr(n; i, j). Therefore
Pn (t) = 'E.i,j rr(n; i, j)JP>(A(t) = i)JP>(B(t) = j).
6.
With p
= A/ p,, the stationary distribution of the imbedded chain is, as in Exercise (11.2.1 ),
{ io  p)
Jrn =
iO  p2)pn1
~
if n = 0,
if n::: 1.
Now'.; rr;
Jrn
J....
and gn =
J....
+ p, for n :::
1, whence,
2
1 for n::: 1.
2 (/...: p,) (1 p )pn
7. (a) Let Q; (t) be the number of people in the ith queue at timet, including any currently in service.
The process Q1 is reversible in equilibrium, and departures in the original process correspond to arrivals
in the reversed process. It follows that the departure process of the first queue is a Poisson process
with intensity J...., and that the departure process of Q 1 is independent of the current value of Q1
(b) We have from part (a) that, for any given t, the random variables Q1 (t), Q2(t) are independent.
Consider an arriving customer when the queues are in equilibrium, and let W; be his waiting time
(before service) in the ith queue. With T the time of arrival, and recalling Exercise (11.2.3),
JP>(W1
= 0,
W2
= 0)
> JP>(Q;(T)
= (1 
P1)(1  P2)
Therefore W1 and W2 are not independent. There is a slight complication arising from the fact that T
is a random variable. However, T is independent of everybody who has gone before, and in particular
of the earlier values of the queue processes Q;.
383
[11.3.1][11.4.1]
Solutions
Queues
where An is the number of arrivals during the (n + l)th service period, and h(m) = 1  8mo Now
Qn and Qn+l have the same distribution. Take expectations to obtain
> 0)
0)}
From the standard theory, Ms satisfies Ms(s) = Ms(s  A. + A.Ms(s)), where Ms(e)
~t/(Jt e). Substitute to find that x = Ms (s) is a root of the quadratic A.x 2  x(A. + 1t s) + 1t = 0.
For some small positive s, M B (s) is smooth and nondecreasing. Therefore M B (s) is the root given.
2.
3. Let Tn be the instant of time at which the server is freed for the nth time. By the lackofmemory
property of the exponential distribution, the time of the first arrival after Tn is independent of all
arrivals prior to Tn, whence Tn is a 'regeneration point' of the queue (so to say). It follows that the
times which elapse between such regeneration points are independent, and it is easily seen that they
have the same distribution.
ao
Trn
= L ai:rrn+i1
for n 2:: 1.
i=O
It is easily seen, by adding, that the first equation is a consequence of the remaining equations, taken
in conjunction with L:Q" :rri = 1. Therefore :rr is specified by the equation for :rrn, n 2:: 1.
384
Solutions [11.4.2][11.5.2]
G/G/1
It is easily seen that A(8) = Mx(~t(8  I)) is convex and nondecreasing on [0, 1], and satisfies
A(O) > 0, A(l) = 1. Now A'(l) = ~tlE(X) = p 1 > 1, implying that there is a unique 'f/ E (0, 1)
such that A('f/) = 'f'J. With this value of 'f'J, the vector 1r given by Trj = (1  TJ)'f/1, j ;:: 0, is a
stationary distribution of the imbedded chain. This 1r is the unique such distribution because the chain
is irreducible.
2. (i) The equilibrium distribution is Trn = (1 TJ)'f'Jn for n 2: 0, with mean l:~o nnn = TJ/0 TJ).
(ii) Using the lackofmemory property of the service time in progress at the time of the arrival, we
see that the waiting time may be expressed as W = S1 + S2 + + S Q where Q has distribution 1r,
given above, and the Sn are service times independent of Q. Therefore
JE(W)
= lE(S1)lE(Q) = ___!}_j__.
(1 'f/)
3.
Mw(s)=
s~t( 1 0 +(1S),
~tOns
the moment generating function of a mixture of an atom at 0 and an exponential distribution with
parameter ~t(1  S).
If G is the probability generating function of the equilibrium queuelength, then, using the lackofmemory property of the exponential distribution, we have that Mw(s) = G(~t/(Jt s)), since W is
the sum of the (residual) service times of the customers already present. Set u = ~t/(Jt s) to find that
G(u) = (1 0/(1 su), the generating function of the mass function f(k) = (1 Osk fork::: 0.
385
[11.5.3](11.7.2]
Solutions
Queues
It may of course be shown that ~ is the smallest positive root of the equation x = M x (~t(x where X is a typical interarrival time.
3.
1)),
We have that
1 G(y)
= lP'(S X>
y)
lx:;
lP'(S > u
+ y)dFx(u),
JR,
where Sand X are typical (independent) service and interarrival times. Hence, formally,
dG(y) =  {
lo
00
dlP'(S > u
+ y) dFx(u)
= dy
1y ~teJJ,(u+y)
00
dFx(u),
x F(x y)dG(y) = {{
{I
oo
}}oo<y::::X
y<u<OO
7JeJJ,(l1})(xy)}~teJJ,(u+y) dFx(u)dy.
First integrate over y, then over u (noting that F x (u) = 0 for u < 0), and the double integral collapses
to F(x), when x ::::: 0.
asptl.
The limit characteristic function is that of the exponential distribution, and the result follows by the
continuity theorem.
n(n) =
ni a
a. e
II 'n'
'
1 
i=I
the product of Poisson distributions. This is related to Bartlett's theorem (see Problem (8.7.6)) by
defining the state A as 'being in station i at some given time'.
2. The number of customers in the queue is a birthdeath process, and is therefore reversible in
equilibrium. The claims follow in the same manner as was argued in the solution to Exercise (11.2.7).
386
Solutions
Problems
[11.7.3][11.8.1]
3. (a) We may take as state space the set {0, 11, 111 , 2, 3, ... }, where i E {0, 2, 3, ... } is the state
of having i people in the system including any currently in service, and 11 (respectively 111 ) is the
state of having exactly one person in the system, this person being served by the first (respectively
second) server. It is straightforward to check that this process is reversible in equilibrium, whence the
departure process is as stated, by the argument used in Exercise (11.2.7).
(b) This time, we take as state space the set {0', 0 11 , 11 , 1", 2, 3, ... } having the same states as in part
(a) with the difference that O' (respectively O") is the state in which there are no customers present and
the first (respectively second) server has been free for the shorter time. It is easily seen that transition
from 01 to 111 has strictly positive probability whereas transition from 111 to 0 1 has zero probability,
implying that the process is not reversible. By drawing a diagram of the state space, or otherwise, it
may be seen that the timereversal of the process has the same structure as the original, with the unique
change that states 01 are 011 are interchanged. Since departures in the original process correspond to
arrivals in the timereversal, the required properties follow in the same manner as in Exercise (11.2.7).
4. The total time spent by a given customer in service may be expressed as the sum of geometrically
distributed number of exponential random variables, and this is easily shown to be exponential with
parameter 8~t. The queue is therefore in effect aM(A.)/M(8~t)ll system, and the stationary distribution
is the geometric distribution with parameter p = A./(8~t), provided p < 1. As in Exercise (11.2.7),
the process of departures is Poisson.
Assume that rejoining customers go to the end of the queue, and note that the number of customers
present constitutes a Markov chain. However, the composite process of arrivals is not Poisson, since
increments are no longer independent. This may be seen as follows. In equilibrium, the probability of
an arrival of either kind during the time interval (t, t +h) is A.h + P~tO 8)h +o(h) = (A.j8)h + o(h).
If there were an arrival of either kind during (t h, t), then (with conditional probability 1  O(h))
the queue is nonempty at time t, whence the conditional probability of an arrival of either kind during
(t, t +h) is A.h + ~tO  8)h + o(h); this is of a larger order of magnitude than the earlier probability
(A./8)h + o(h).
5. For stations r, s, we write r + s if an individual at r visits s at a later time with a strictly positive
probability. Let C comprise the station j together with all stations i such that i + j. The process
restricted to C is an open migration process in equilibrium. By Theorem (11. 7 .19), the restricted
process is reversible, whence the process of departures from C via j is a Poisson process with some
intensity t;. Individuals departing C via j proceed directly to k with probability
Ajk~j(nj)
Ajk
1tj~j(nj)+l:.rrtCAjr~j(nj) =
1tj+l:.rrtCAjr'
1::::; n < N,
Jt'lrN + A'lrN1 = 0,
N
'Jf1 ="'\:'(A./ )n =
L
n=O
1t
387
(A./ )N+1
1t
1(A/~t)
=f. Jt)
[11.8.2][11.8.3]
Solutions
Queues
A.
0
ifi < N,
ifi 2:: N,
/Li =
/L
2JL
ifi = 1,
ifi 2:: 2.
/Li+lni+l We deduce
no'= 1 + 2.:2pi.
i=l
2. The answer is obtainable in either case by following the usual method. It is shorter to use the fact
that such processes are reversible in equilibrium.
(a) The stationary distribution1r satisfies nnA.p(n) = nn+l/L for n 2:: 0, whence nn = nopn fn! where
p = A./JL. Therefore nn = pneP fn!.
(b) Similarly,
n1
n 2::0,
nn = nop n
P m = nop n 2  ~n(n1)
",
II ( )
m=O
where
oo
no'= LPn(~)Zn(n1).
I
n=O
At the instant of arrival of a potential customer, the probability q that she joins the queue is
obtained by conditioning on its length:
oo
oo
oo
1
=2.:
p(n)nn =no 2.: pn2nzn<n1) =no 2.: pn2zn<n+l) =no{no' 1}.
I
n=O
n=O
n=O
3. First method. Let (Ql, Qz) be the queuelengths, and suppose they are in equilibrium. Since
Q, is a birthdeath process, it is reversible, and we write (h(t) = Q,(t). The sample paths of
Q, have increasing jumps of size 1 at times of a Poisson process with intensity A.; these jumps mark
arrivals at the cash desk. By reversibility, Q, has the same property; such increasing jumps for Q,
are decreasing jumps for Q 1, and therefore the times of departures from the cash desk form a Poisson
process with intensity A.. Using the same argument, the quantity Q, (t) together with the departures
prior tot have the same joint distribution as the quantity Q, (t) together with all arrivals after t.
However Q, (t) is independent of its subsequent arrivals, and therefore Q, (t) is independent of its
earlier departures.
It follows that arrivals at the second desk are in the manner of a Poisson process with intensity
A., and that Qz(t) is independent of Q, (t). Departures from the second desk form a Poisson process
also.
Hence, in equilibrium, Q, is M(A.)/M(J.LI)/1 and Qz is M(A.)/M(J.Lz)/1, and they are independent
at any given time. Therefore their joint stationary distribution is
nmn =IP'(Q,(t) =m, Qz(t) =n) = (1p,)(lpz)PiP!J.
= A./ J.Li.
Second method. The pair (Ql (t), Qz(t)) is a bivariate Markov chain. A stationary distribution
(nmn : m, n 2:: 0) satisfies
where Pi
= A.nml,n +
J.Linm+l,n1 + J.Lznm,n+l
388
m,n 2::1,
Solutions [11.8.4][11.8.5]
Problems
together with other equations when m = 0 or n = 0. It is easily checked that these equations have the
solution given above, when Pi < 1 for i = 1, 2.
4. Let Dn be the time of then th departure, and let Q n = Q ( Dn +) be the number of waiting customers
immediately after Dn. We have in the usual way that Qn+l = An+ Qn  h(Qn), where An is the
number of arrivals during the (n + l)th service time, and h(x) = min{x, m}. Let G(s) = l:~o :rr;si
be the equilibrium probability generating function of the Qn. Then, since Qn is independent of An,
where
lE(sAn)
fooo
e)..u(sl) fs(u)
du
= Ms (A.(s
1)),
smG(s) = Ms(A.(s
1)) { G(s) +
fcsm  si):rr; }
1=0
whenever it exists.
Finally suppose that m = 2 and Ms(e) = M/(M 8). In this case,
M{rro(s + 1) + :rr1s}
G(s) = ''=='M(s + 1) A.s 2
Now G(l) = 1, whence M(2:rro + :rr1) = 2M A.; this implies in particular that 2M A. > 0. Also
G(s) converges for Is I _:::: 1. Therefore any zero of the denominator in the interval [1, 1] is also a
zero of the numerator. There exists exactly one such zero, since the denominator is a quadratic which
takes the value A. at s = 1 and the value 2M  A. at s = 1. The zero in question is at
= 0.
1a
as
G(s) = 1  ,
where a= 2A./{M +
5.
J M2 + 4A.J[}.
V(A. +
2)..
389
M s) 2  4A.J[
[11.8.6][11.8. 7]
Solutions
Queues
Now M B (s) is nondecreasing in s, and therefore it is the value with the minus sign. The density
function of B may be found by inverting the moment generating function; see Feller (1971, p. 482),
who has also an alternative derivation of M B.
As for the mean and variance, either differentiate MB, or differentiate(*). Following the latter
route, we obtain the following relations involving M (= MB):
2AM M 1 + M
+ (s A 
JL)M 1
= 0,
Sets = 0 to obtain M 1 (0) = (JL A) 1 and M 11 (0) = 2J.L(J.LA) 3 , whence the claims are immediate.
6. (i) This question is closely related to Exercise (11.3.1). With the same notation as in that solution,
we have that
where h(x) = min{l, x}. Taking expectations, we obtain IP'(Qn > 0) = JE(An) where
lE(An) =
lX>
and S is a typical service time. Square (*) and take expectations to obtain
p(l 2p)
lE(Qn) =
+ JE(A~+l)
2 (1 _ p)
Subtract F(x; t), divide by h, and take the limit ash ,).. 0, to obtain the differential equation.
We take LaplaceStieltjes transforms. Integrating by parts, for e :::: 0,
{
lco,oo)
{
lco,oo)
{
lco,oo)
eex dlP'(U
+ S:::: x) =
Mu(B)Ms(B),
and therefore
0 = h(O) B{Mu(B) H(O)}
390
Solutions [11.8.8][11.8.10]
Problems
1) e}.
Take the limit as e .. 0, using L'Hopital's rule, to obtain H(O) = 1 'AJE.(S) = 1  p. The moment
generating function of U is given accordingly. Note that Mu is the same as the moment generating
function of the equilibrium distribution of actual waiting time. That is to say, virtual and actual
waiting times have the same equilibrium distributions in this case.
8. In this case U takes the values 1 and 2 each with probability ~ (as usual, U = S X where S
and X are typical (independent) service and interarrival times). The integral equation for the limiting
waiting time distribution function F becomes
F(O) = ~ F(2),
F(x)
J5).
l: JSr
for some constants A and B. Now F(x) > 1 as x > oo, since the queue is stable, and therefore
A= 1. Using the equation for F(O), we find that B = ~(1
JS).
+ p(l eJLt),
where At has the Poisson distribution with parameter A. When the queue is in equilibrium, JE.(Q(t)) =
+ 1)), and hence
JE.(Q(t
391
+ lP'(Q =
0) }.
[11.8.11][11.8.13]
Solutions
Queues
Also,
G(s) = JE.(sQ I{Q:::1J) + JP>(Q = 0),
and hence
G(s) =
whence
e).(s 1) {
~G(s) +
( 1
~) (1 A)}
(1  s)(l A)
G(s) = 1 se ).(s 1).
The mean queue length is G' (1) = ~A(2  A)/(1 A). Since service times are of unit length,
and arrivals form a Poisson process, the mean residual service time of the customer in service at an
arrival time is ~, so long as the queue is nonempty. Hence
A
11. The length B of a typical busy period has moment generating function satisfying M B (s) =
exp{s A+ AMs(s)}; this fact may be deduced from the standard theory ofM/G/1, or alternatively
by a randomwalk approach. Now T may be expressed as T = I + B where I is the length of the
first idle period, a random variable with the exponential distribution, parameter A. It follows that
MT(s) = AMs(s)/(A s). Therefore, as required,
(A s)MT(s) = Aexp{s A+ (A s)MT(s) }.
If A :::: 1, the queue~length at moments of departure is either null persistent or transient, and it
follows that JE.(T) = oo. If A < 1, we differentiate(*) and sets = 0 to obtain AlE.(T) 1 =A 2 JE.(T),
whence JE.(T) = {A(l  A)} 1.
12. (a) Q is a birthdeath process with parameters Ai = A, /[i = {[, and is therefore reversible in
equilibrium; see Problems (6.15.16) and (11.8.3).
(b) The equilibrium distribution satisfies krci = {[7ri+1 fori :::: 0, whence Jri = (1 p)pi where
p = A/ f.[. A typical waiting time W is the sum of Q independent service times, so that
Mw(s)=GQ(Ms(s))=
1p
1 P{[/({[ s)
(1p)(f.[S)
.
f.[(1 p) s
where
Pj
=A/{[J.
13. The size of the queue is a birthdeath process with rates Ai =A, /[i = {[ min{i, k}. Either solve
the equilibrium equations in order to find a stationary distribution 7r, or argue as follows. The process
is reversible in equilibrium (see Problem (6.15.16)), and therefore Ailri = /[i+17ri+1 for all i. These
'balance equations' become
ifO~i<k,
ifi:::: k.
392
Solutions [11.8.14][11.8.14]
Problems
Tri
noai I i!
no(aj k)i kk I k!
ifO:Si ::Sk,
ifi:::: k
where a = A./ JL. Therefore there exists a stationary distribution if and only if J.. . < kJL, and it is given
accordingly, with
c1
Ba
=A+,
1a
2Ba 2
4a
C2 =2A+  2.
Therefore
(1  a)(4 a2)
Viewed as a function of a, the numerator is a cubic taking the value 4A at a= 0 and the value 3B
at a= l. This cubic has a unique zero at some a* E (0, 1), and C1 < C2 if and only ifO <a <a*.
14. The state of the system is the number Q(t) of customers within it at timet. The state 1 may be
divided into two substates, being a1 and a2, where O"i is the state in which server i is occupied but
the other server is not. The state space is therefore S = {0, a1, a2, 2, 3, ... }.
The usual way of finding the stationary distribution, when it exists, is to solve the equilibrium
equations. An alternative is to argue as follows. If there exists a stationary distribution, then the
process is reversible in equilibrium if and only if
for all sequences i1, i2, ... , ik of states, where G = (guv)u,vES is the generator of the process (this
may be shown in very much the same way as was the corresponding claim for discretetime chains
in Exercise (6.5.3); see also Problem (6.15.16)). It is clear that(*) is satisfied by this process for all
sequences of states which do not include both a1 and a2; this holds since the terms guv are exactly
those of a birthdeath process in such a case. In order to see that (*) holds for a sequence containing
both a1 and a2, it suffices to perform the following calculation:
go,al gal ,2g2,azgaz,O = ( ~ J....)J....J.L2/Ll = go,az gaz,2g2,al gal ,0
1r:
satisfies Truguv
and hence
Tra1
= 2
J.. .
ILl
no,
')...2
393
')...
JL1
+ JL2
)u2 no
for u ::=: 2.
[11.8.15][11.8.17]
Solutions
Queues
This gives a stationary distribution if and only if).. < JL1 + f.i2, under which assumption no is easily
calculated.
A similar analysis is valid if there are s servers and an arriving customer is equally likely to go
to any free server, otherwise waiting in tum. This process also is reversible in equilibrium, and the
stationary distribution is similar to that given above.
15. We have from the standard theory that QJL has as mass function Trj = (1  IJ)IJj, j 0::: 0, where
IJ is the smallest positive root of the equation x = eJL(xI). The moment generating function of
(1 f.L 1)QJL is
1 1] _ .
MJL(8)=E(exp{8Clf.Ll)QJL})=
1
1  IJeeOJL l
Writing f.L = 1 + E, we have by expanding eiLC'7l) as a Taylor series that IJ = IJ(E) = 1 2E + o(E)
as E ,j, 0. This gives
2E + O(E)
2E + O(E)
2
= 1  (1  2E)(l +BE)+ o(E) = (2 B)E + o(E) + 2 e
MJL(B)
16. The numbers P (of passengers) and T (of taxis) up to timet have the Poisson distribution with
respective parameters nt and rt. The required probabilities Pn = IP'(P = T + n) have generating
function
00
00
n=OOm=O
00
= GT(Zl)Gp(Z) = e(rr+r)te(rrz+rzl)t,
in which the coefficient of zn is easily found to be that given.
17. Let N(t) be the number of machines which have arrived by timet. Given that N(t) = n, the times
T1, T2, ... , Tn of their arrivals may be thought of as the order statistics of a family of independent
uniform variables on [0, t], say U1, U2, ... , Un; see Theorem (6.12.7). The machine which arrived
at time
ui
is, at time t,
in the X stage}
{ a(t)
in the Y stage
with probability
f3 ( t)
1  a(t)  f3(t)
repaired
where a(t) = IP'(U +X > t) and f3(t) = IP'(U +X :=: t < U +X+ Y), where U is uniform on [0, t],
and (X, Y) is a typical repair pair, independent of U. Therefore
IP'(U(t) = . V(t) = k I N(t) = n) = n! a(t)j f3(ti (1  a(t)  f3(t))nk j
1"lkl( n 1k)'
implying that
oo eAt (At)n
IP'(U(t) = j, V(t) = k) =
2.:
n= 0
n.1
.,
394
k!
Solutions [11.8.18][11.8.19]
Problems
18. The maximum deficit Mn seen up to and including the time of the nth claim satisfies
Mn
where the Xj are the interclaim times, and Uj = Kj  Xj. We have as in the analysis of G/G/1 that
Mn has the same distribution as Vn = max{O, Un, Un + Un1, ... , Un + Un1 + + Ul}, whence
Mn has the same distribution as the (n + l)th waiting time in a M(A.)/G/1 queue with service times
Kj and interarrival times Xj. The result follows by Theorem (11.3 .16).
19. (a) Look for a solution to the detailed balance equations A.rri = (i + l)JL7ri+1, 0
that the stationary distribution is given by Jri = (pi/ i !)no.
(b) Let Pc be the required fraction. We have by Little's theorem (10.5.18) that
Pc =
A.(rrc1  Jrc)
= p(rrc1  rrc),
i < s, to find
c :=:: 2,
fL
and P1 = rr1, where Jrs is the probability that channels 1, 2, ... , s are busy in a queue M/M/s having
the property that further calls are lost when all s servers are occupied.
395
12
Martingales
(i) We have that E(Ym) = E{E(Ym+l I Tm)} = E(Ym+J), and the result follows by induction.
(ii) For a submartingale, E(Ym) ::0 E{E(Ym+l I Tm)} = E(Ym+d, and the result for supermartingales
follows similarly.
2.
We have that
if m :::: 1, since Tn s; Tn+m1 Iterate to obtain E(Yn+m I Tn) = E(Yn I Tn) = Yn.
3.
where the Xi are independent family sizes with probability generating function G. Now G(ry) = ry,
and the claim follows.
4.
where Tn = cr(X 1, Xz, ... , Xn). Also EISn I ::0 n, so that {Sn} is a martingale.
(ii) Similarly E(S;) = var(Sn) = n, and
(iii) Suppose the walk starts at k, and there are absorbing barriers at 0 and N (:::: k). LetT be the time
at which the walk is absorbed, and make the assumptions that E(ST) =So, E(Sf T) =
Then
the probability Pk of ultimate ruin satisfies
sg.
0 Pk
+N
(1  Pk) = k,
396
Solutions
Introduction
[12.1.5][12.1.9]
= lE{lE(YrYi
I .rj)}
= lE{YilE(Yr
I .rj)}
= lE(Yh,
+ lE(Y}I
.rj ).
j.:::: k.
Now {lE(YJ) : n :;:: 1} is nondecreasing and bounded, and therefore converges. Therefore, by(*),
{ Yn : n :;:: 1} is Cauchy convergent in mean square, and therefore convergent in mean square, by
Problem (7.11.11).
6.
(ii) It suffices to note that lx I, x 2 , and x+ are convex functions of x; draw pictures if you are in doubt
about these functions.
7. (i) This follows just as in Exercise (12.1.6), using the fact that u{lE(Yn+l I .7='n)} ::=: u(Yn) in this
case.
(ii) The function x+ is convex and nondecreasing. Finally, let {Sn : n ::=: 0} be a simple random walk
whose steps are +1 with probability p (= 1  q > ~) and 1 otherwise. If Sn < 0, then
lE (ISn+III .7='n)
= p(ISnl1) + q(ISnl + 1) =
note that lP'(Sn < 0) > 0 if n ::=: 1. The same example suffices in the remaining case.
8.
S}. Also,
where .7='n = cr(X 1, Xz, ... , Xn). Divide by A. n+l to obtain that the given sequence is a supermartingale.
9. Since var(ZJ) > 0, the function G, and hence also Gn, is a strictly increasing function on [0, 1].
Since 1 = Gn+l (Hn+l (s)) = Gn(G(Hn+l (s))) and Gn(Hn(s)) = 1, we have that G(Hn+l (s)) =
Hn(s). With .7='m = cr(Zk : 0 _:::: k _:: : m),
lE(Hn+l (s) 2 n+l I .7='n) = G(Hn+l (s))Zn = Hn(s)Zn.
397
[12.2.1][12.3.2]
Solutions
Martingales
I:FJ) =
lE(Z(j)
I0d.
+ M.
Take conditional expectations of the second inequality, given :F) and given :F) I , and deduce that
IYj  YJ!1 _::: M. Therefore Y is a martingale with bounded differences, and Hoeffding's inequality
yields the result.
2. Let :fi be the afield generated by the (random) edges joining pairs (va, Vb) with 1 _::: a, b _::: i,
and let Xi = lE(X I :fi ). We write x (j) for the minimal number of colours required in order to colour
each vertex in the graph obtained by deleting Vj. The argument now follows that of the last exercise,
using the fact that X (j) _::: X _::: X (j) + 1.
Let T1
= min{n:
Yn
:=:::
b}, Tz
= min{n
b},
Zn =
L li(Yi YiJ),
:=:::
0.
i=l
+ (Yn b)+,
whence
Ui
= 1} =
Therefore,
since In :=::: 0 andY is a submartingale. It follows that lE(Zn) :=::: lE(Zn_J) :=::: :=::: lE(Zo) = 0, and
the final inequality follows from ( *).
2. If Y is a supermartingale, then  Y is a submartingale. Upcrossings of [a, b] by Y correspond to
downcrossings of [ b, a] by  Y, so that
lEUn (a, b, Y) =lEDn (b ,  a, Y) <

398
lE{( Y +a)+}
n
ba
JE{(Y a)}
= __bn_ __
a
,
Solutions
Stopping times
[12.3.3][12.4.5]
I:i"'
4. Y is a martingale since Yn is the sum of independent variables with zero means. Also
IP'(Zn =f.
0) = L:j"' n 2 < oo, implying by the BorelCantelli lemma that Zn = 0 except for finitely many
values of n (a.s.); therefore the partial sum Yn converges a.s. as n + oo to some finite limit.
It is easily seen that an = San 1 and therefore an = 8 sn 2 , if n :=:: 3. It follows that IYn I :=:: ian
if and only if IZn I = an. Therefore
We have that
n
U ({T1=k}n{T2=nk}),
k=O
{ max{T1, T2}::::: n} = m : : : n} n {T2::::: n},
{ min{T1, T2}::::: n} = m : : : n} U {T2::::: n}.
{T1+T2=n}=
2.
+1=
n} = {Sn1::::: t}
n {Sn
> t}
:Fn.
3. (Y+, :F) is a submartingale, and T = min{k: Yk :=:: x} is a stopping time. Now 0::::: T 1\ n::::: n,
so that JE(Yd) ::::: lE(YfAn) ::::: lE(Y,i), whence
4.
We may suppose that JE(Yo) < oo. With the notation of the previous solution, we have that
5. It suffices to prove that lEYs ::::: lEYT, since the other inequalities are of the same form but with
different choices of pairs of stopping times. Letlm be the indicator function of the event {S < m ::::: T},
and define
n
Zn =
lm(Ym Ym1),
m=1
399
O:sn :S N.
[12.4.6][12.5.2]
Solutions
Martingales
~ E(Zo)
= 0.
~ x)
= ll' (max Ym
O:=;m:=;n
~ (qlp)x)::: (plq)x.
L ll'(S
E(Soo) =
00
x=l
x)::: _P__
q P
We can calculate E(S00 ) exactly as follows. It is the case that S00 ~ x if and only if the walk
ever visits the point x, an event with probability fx for x ~ 0, where f = pI q (see Exercise (5.3.1)).
The inequality of (*) may be replaced by equality.
7.
(a) First,
n {T::: n} =
0 E :Fn.
Secondly, if An {T::: n}
E :Fn
then
n {T ::: n} E
{T<m}n{T<n}= {
{T:Sm}
an event lying in :Fn. Therefore {T ::: m}
(b) Let A E :Fs. Then, for any n,
E :Fr
ifm > n,
ifm::: n,
for all m.
(An{S:::T})n{T:Sn}=
U (An{S:::m})n{T=m},
m=O
the union of events in :Fn, which therefore lies in :Fn. Hence A n {S ::: T} E :Fr.
(c) We have {S::: T} = n, and (b) implies that A E :Fr whenever A E :Fs.
1.
400
0} is
Solutions
Optional stopping
[12.5.3][12.5.5]
(a) Now,
lE (IYTAnll(IYT/\nl:::aJ) = lE (IYrll(T:sn,IYTI2:aJ) + lE (IYnll(T>n,IYnl:::aJ)
::0 lE (IYrll(lhl:::aJ) + lE (IYnll{T>nJ) = g(a) + h(n),
say. We have that g(a) + 0 as a + oo, since lEIYrl < oo. Also h(n) + 0 as n + oo,
so that supn>N h(n) may be made arbitrarily small by suitable choice of N. On the other hand,
lE (IYnll(IYnl~aJ)+ 0 as a+ oo uniformly inn E {0, 1, ... , N}, and the claim follows.
(b) Since Y;i defines a submartingale, we have that supn lE(YfAn) ::0 supn JE(Y;i) < oo, the second
inequality following by the uniform integrability of {Yn }. Using the martingale convergence theorem,
YT/\n+ Yr a.s. wherelEIYrl < oo. Now
Also lP'(T > n) + 0 as n + oo, so that the final two terms tend to 0 (by the uniform integrability of
the Yi and the finiteness of lEI Yr I respectively). Therefore Yr An
standard theorem (7.10.3).
3. By uniform integrability, Y00 = limn+oo Yn exists a.s. and in mean, and Yn = lE(Y00 I :Fn).
(a) On the event {T = n} it is the case that Yr = Yn and JE(Y00 I :Fr) = lE(Y00 I :Fn); for the
latter statement, use the definition of conditional expectation. It follows that Yr = JE(Y00 I :Fr ),
irrespective of the value of T.
(b) We have from Exercise (12.4.7) that :Fs ~ :Fr. Now Ys = JE(Yoo I :Fs) = lE{lE(Y00 I :Fr) I
:Fs)
= JE(Yr I :Fs).
4.
Let T be the time until absorption, and note that {Sn} is a bounded, and therefore uniformly
integrable, martingale. Also IP'(T < oo) = 1 since T is no larger than the waiting time for N
consecutive steps in the same direction. It follows that JE(So) = lE(Sr) = NIP'(Sr = N), so that
IP'(Sr = N) = JE(So)/ N. Secondly, {S~ n : n 2':: 0} is a martingale (see Exercise (12.1.4)), and the
optional stopping theorem (if it may be applied) gives that
lE(S6)
= JE(Sf 
T)
= N 21P'(Sr = N) lE(T),
5. Let :Fn = cr(SJ, S2, ... , Sn). It is immediate from the identity cos(A + ).) + cos(A ).) =
2 cos A cos). that
lE(Yn+l I :Fn)
2(cos ,J,.)n+l
= Yn,
and therefore Y is a martingale (it is easy to see that lEI Yn I < oo for all n ).
SupposethatO <). < rrj(a +b), andnotethatO ::0 I,J,.{Sn ~(b a)}l <~).(a +b)< ~JT for
n ::0 T. Now Yr An constitutes a martingale which satisfies
cos{~,J,.(a+b)}
_ _=c__=_ < Yr An <
(cos ,J,.)T 1\n
401
(cos ,J,.)T
[12.5.6][12.5.8]
Solutions
Martingales
If we can prove that E{(cos A.)T} < oo, it will follow that {Yr 1\n} is uniformly integrable. This will
imply in tum that E(Yr) = Iimn+oo E(Yr 1\n) = E(Yo), and therefore
cos{~A.(a + b)}E{(cosA.)T} = cos{iA.(b a)}
Now T
1\
E{(cosA.)T}:::
cos{iA.(a b)}
E(Yo)
cos{~A.(a +b)}
cos{iA.(a +b)}
6. (a) The occurrence of the event {U = n} depends on Sr, S2, ... , Sn only, and therefore U is a
stopping time. Think of U as the time until the first sequence of five consecutive heads in a sequence
of coins tosses. Using the renewaltheory argument of Problem (10.5.17), we find that E(U) = 62.
(b) Knowledge of Sr, S2, ... , Sn is insufficient to determine whether or not V = n, and therefore V
is not a stopping time. Now E(V) = E(U) 5 =57.
(c) W is a stopping time, since it is a firstpassage time. Also E(W) = oo since the walk is null
persistent.
7.
E(Mm+n I :Fm)
=E(
r=O
Sr
Sr 
~(Sm+n 
Sm
+ Sm) 3
:Fm)
r=m+l
= Mm
Thus {Mn : n :::: 0} is a martingale, and evidently Tis a stopping time. The conditions of the optional
stopping theorem (12.5.1) hold, and therefore, by a result of Example (3.9.6),
8. We partition the sequence into consecutive batches of a + b flips. If any such batch contains only
l's, then the game is over. Hence lP'(T > n(a +b))::; {1 (~)a+b}n+ 0 as n+ oo. Therefore,
EISf Tl ::; E(Sf) + E(T) ::; (a+ b) 2 + E(T) < oo,
and
as n+ oo.
402
Solutions
Problems
[12.7.1][12.9.2]
Lets~ t. WehavethatE(I](X(t))
I J=S, Xs =
i)
:tlE(IJ(X(t)) :Fs, Xs =
= (PrsGJ7'); = 0,
so that E(I](X (t)) I :Fs, Xs = i) = l](i ), which is to say that lE(IJ (X (t))
e 0::: 0.
1\
Ta), t 0::: 0,
where, by assumption, the limit has finite expectation for sufficiently small positive e (this fact may be
checked easily). In this case, {W(t 1\ Ta): t 0::: 0} is uniformly integrable. Now W(t 1\ Ta)+ W(Ta)
a.s. as t + oo, and it follows by the optional stopping theorem that
Writes = eli to obtain sa = E{eHae 1sl}. Differentiate at s = 1 to find that a = AE(Ta) and
a(a + 1) =A 2E(T}), whence the claim is immediate.
3. Let 9m be the a field generated by the two sequences of random variables Sm, Sm+ 1 ... , Sn and
Um+1, Um+2 ... , Un. It is a straightforward exercise in conditional density functions to see that
1
E (Um+1
!oUm+2
I 9m+1)  0
(m
+ l)xm 1 d x m+ 1
(U
m+2
)m+1
U
'
m m+2
~ = E(RT I Sn =
t
~ n I Sn = y).
[Equality may be shown to hold. See Karlin and Taylor 1981, pages 110113, and Example (12.7.6).]
Clearly E(Zn)
'f,:; 1 X;+ A, where X1, X2, ... are the family sizes of the members of the nth generation, and A is
the number of immigrants to the (n + l)th generation. Therefore E(Zn+1
I Zn) = tJZn +
m, whence
1 { MZn + m ( 1 1 l _Mn+1)}
E(Yn+1 I Zn) = Mn+
M
= Yn.
1
2. Each birth in the (n + 1)th generation is to an individual, say the sth, in the nth generation. Hence,
for each r, Ben+ 1),r may be expressed in the form Ben+ 1),r = Bn,s + Bj(s ), where Bj (s) is the age
of the parent when its jth child is born. Therefore
E{
S,J
403
[12.9.3][12.9.5]
Solutions
Martingales
1.
If x, c > 0, then
lP' (max Yk >
1~k~n
1~k~n
2 ).
Now (Yk + c) 2 is a convex function of Yk, and therefore defines a submartingale (Exercise (12.1.7)).
Applying the maximalinequality to this submartingale, we obtain an upper bound ofE{ (Yn +c) 2 }/ (x +
c) 2 for the righthand side of(*). We set c = E(Y1)/x to obtain the result.
4. (a) Note that Zn = Zn1 + Cn{Xn E(Xn I :Fn_I)}, so that (Z, !F) is a martingale. LetT be
the stopping timeT= min{k: qYk 2:: x}. ThenE(ZTAn) = E(Zo) = 0, so that
k=1
where we have used the facts that Yn 2:: 0 and E(Xk I :Fkd 2:: 0. The claim follows.
(b) Let X 1, X 2, ... be independent random variables, with zero means and finite variances, and let
Yj =
1 X;. Then Y/ defines a nonnegative submartingale, whence
L;{=
5. The function h(u) = lulr is convex, and therefore Y;(m) = IS; Smlr, i 2:: m, defines a
submartingale with respect to the filtration :F; = a ({x1 : 1 :::: j :::: i}). Apply the HRC inequality of
Problem (12.9.4), with q = 1, to obtain the required inequality.
If r = 1, we have that
m+n
L
EIZkl
k=m+1
by the triangle inequality. Let m, n + oo to find, in the usual way, that the sequence {Sn} converges
a.s.; Kronecker's lemma (see Exercise (7.8.2)) then yields the final claim.
Suppose 1 < r ::=: 2, in which case a little more work is required. The function h is differentiable,
and therefore
Now h'(y) = rlylr 1sign(y) has a derivative decreasing in IYI It follows (draw a picture) that
h 1 (u + x) h 1(u) ::=: 2h 1 ( ix) if x 2:: 0, and therefore the above integral is no larger than 2h(i(v u)).
Apply this with v = Sm+k+1  Sm and u = Sm+k Sm, to obtain
404
Solutions [12.9.6][12.9.10]
Problems
to deduce that
E(ISm+n smn
~ 2 2 r
m+n
E(lzkn
k=m+l
With
0 if and only if Xn
0. Therefore lP'(Yn
+ niYntl)}
=
0)
lP'(Xn
0)
1  n 1 + 1
as n + oo, implying that Yn ~ 0. On the other hand, Ln lP'(Xn =1 0) = oo, and therefore
lP'(Yn =1 0 i.o.) = 1 by the second BorelCantelli lemma. However, Yn takes only integer values, and
therefore Yn does not converge to 0 a.s. The martingale convergence theorem is inapplicable since
supn EIYnl = 00.
7. Assume that t > 0 and M(t) = 1. Then Yn = e 1Sn defines a positive martingale (with mean 1)
with respect to .Tn = a(Xt, X2, ... , Xn). By the maximal inequality,
8. The sequence Yn = ~Zn defines a martingale; this may be seen easily, as in the solution to
Exercise (12.1.15). Now {Yn} is uniformly bounded, and therefore Y 00 = limn+oo Yn exists a.s. and
satisfies E(Y00 ) = E(Yo) = ~.
Suppose 0 < ~ < 1. In this case Zt is not a.s. zero, so that Zn cannot converge a.s. to a constant
c unless c E {0, oo}. Therefore the a.s. convergence of Yn entails the a.s. convergence of Zn to a limit
random variable taking values 0 and oo. In this case, E(Y00 ) = 1 lP'(Zn + 0) + 0 lP'(Zn + oo ),
implying that lP'(Zn + 0) = ~, and therefore lP'(Zn + oo) = 1  ~.
9. It is a consequence of the maximal inequality that lP'(Y; =:: x) ~ x 1E(Ynl{Y,i::::x)) for x > 0.
Therefore
E(Y;) = fooo lP'(Y; =:: x) dx
~ 1 + E { Yn
00
~ 1+
00
lP'(Y;
=:: x) dx
+ E(Y;)fe.
=i) =
LPij(t)h(j)
j
405
fors < t,
[12.9.11][12.9.13]
Solutions
Martingales
for any event B defined in terms of {X (u) : u ::=: s}. The derivative of this expression, with respect to
t, is (P1 Gh')i, where P1 is the transition semigroup, G is the generator, and h = (h(j) : j :::: 0). In
this case,
(Gh1)j
=L
gjkh(k)
for all j. Therefore the left side of (*) is constant for t ::=: s, and is equal to its value at time s, i.e.
X (s). Hence h(X(t)) defines a martingale.
(b) We apply the optional stopping theorem with T = min{t: X(t) E {0, n}} toobtainE(h(X(T))) =
E(h(X(O))), and therefore (1  n(m))h(n) = h(m) as required. It is necessary but not difficult to
check the conditions of the optional stopping theorem.
Therefore {E(Y:+m I :Fn) : m :::: 0} is (a.s.) nondecreasing, and therefore converges (a.s.) to a limit
Mn. Also, by monotone convergence of conditional expectation,
E(Mn+l
I :Fn) = m+oo
lim E{E(Y:+m+l I :Fn+I) I:Fn} = m+00
lim E(Y:+m+l I :Fn) = Mn,
and furthermore E(Mn) = limm+oo E(Y:+n) :::0 M. It is the case that Mn is :Fnmeasurable, and
therefore it is a martingale.
(b) We have that Zn = Mn  Yn is the difference of a martingale and a submartingale, and is therefore
a supermauingale. Also Mn ::=: Y: ::=: 0, and the decomposition for Yn follows.
(c) In thi~ case Zn is a martingale, being the difference of two martingales. Also Mn :::: E(Y: I :Fn) =
Y: ::=: Yn a.s., and the claim follows.
12. We may as well assume that JJ < P since the inequality is trivial otherwise. The moment
generating function of P Cr is M(t) = iCPJLH5;a 212 , and we choose t such that M(t) = l,
i.e., t = 2(P J.J)/a 2. Now define Zn = min{etYn, 1} and :Fn = a(Cr, C2, ... , Cn). Certainly
EIZnl < oo; also
E(Zn+l I :Fn) :::0 E(etYn+l I :Fn) = etYn M(t) = etYn
and E(Zn+l I :Fn) :S 1, implying that E(Zn+l I :Fn) :S Zn. Therefore (Zn, :Fn) is a positive
supermartingale. LetT = inf{n : Yn ::=: 0} = inf{n : Zn = 1}. Then T 1\ m is a bounded stopping
time, whence E(Zo) ::=: E(ZT Am) ::=: JP>(T ::=: m). Let m + oo to obtain the result.
E(Rn+l
I Rn)
= Rn
Rn
+ n+r+ b'
whence Yn satisfies E(Yn+I I :Fn) = Yn. Therefore {Yn : n ::=: 0} is a uniformly integrable martingale,
and therefore converges a.s. and in mean.
(b) In order to apply the optional stopping theorem, it suffices that JP>(T < oo) = l (since Y is
uniformly integrable). However JP>(T > n) =
~
n~I = (n + l) 1 + 0. Using that theorem,
1
E(YT) = E(Yo), which is to say that E{T I (T + 2)} = 1, and the result follows.
(c) Apply the maximal inequality.
406
Solutions [12.9.14][12.9.17]
Problems
14. As in the previous solution, with fi,n the afield generated by A1, A2, ... and :Fn,
E(Yn+1
I 13
(J>n) =
Rn + An
) (
Rn
)
Rn + Bn +An
Rn + Bn
Rn
=Yn,
Rn +Bn
(
+(
Rn
Rn
+ Bn + An
) (
Bn
Rn
+ Bn
sothatE(Yn+1 I :Fn) = E{E(Yn+1 I fi,n) I :Fn} = Yn. Also IYn I ::=: 1, and therefore Yn is a martingale.
We need to show that lP'(T < oo) = 1. Letln be the indicator function of the event {T > n}. We
have by conditioning on the An that
E(In
I A)
IT
J=O
2 + SJ
IT
00
J=O
1  1 )
2 + S1
L{=
15. At each stage k, let Lk be the length of the sequence 'in play', and let Yk be the sum of its
entries, so that Lo = n, Yo = L,j= 1 Xi. If you lose the (k + 1)th gamble, then Lk+ 1 = Lk + 1 and
Yk+1 = Yk + Zk where Zk is the stake on that play, whereas if you win, then Lk+1 = Lk  2 and
Yk+1 = Yk Zk; we have assumed that Lk :::: 2, similar relations being valid if Lk = 1. Note that
Lk is a random walk with mean step size 1, implying that the firstpassage time T to 0 is a.s. finite,
and has all moments finite. Your profits at time k amount to Yo  Yb whence your profit at time T is
Yo, since YT = 0.
Since the games are fair, Yk constitutes a martingale. Therefore E(YT 1\m) = E(Yo) =j:. 0 for
all m. However T 1\ m + T a.s. as m + oo, so that YT 1\m + YT a.s. Now E(YT) = 0 =j:.
Iimm+ooE(YT/\m), and it follows that {YT/\m : m :0:: 1} is not uniformly integrable. Therefore
E(supm YT 1\m) = oo; see Exercise (7.10.6).
16. Since the game is fair, E(Sn+1 I Sn) = Sn. Also ISnl ::=: 1 + 2 + + n < oo. Therefore Sn is
a martingale. The occurrence of the event {N = n} depends only on the outcomes of the cointosses
up to and including the nth; therefore N is a stopping time.
A tail appeared at time N 3, followed by three heads. Therefore the gamblers G1, G2, ... ,
G N 3 have forfeited their initial capital by time N, while G N i has had i + 1 successful rounds for
0 ::=: i ::=: 2. Therefore SN = N (p 1 + p 2 + p 3 ), after a little calculation. It is easy to check that
N satisfies the conditions of the optional stopping theorem, and it follows that E(SN) = E(So) = 0,
which is to saythatE(N) = p1 + p2 + p3.
In order to deal with HTH, the gamblers are reprogrammed to act as follows. If they win
on their first bet, they bet their current fortune on tails, returning to heads thereafter. In this case,
SN = N (p 1 + p 2q 1) where q = 1 p (remember that the game is fair), and therefore
E(N) = p1 + p2q1.
17. Let :Fn = a ({Xi , Yi : 1 ::=: i ::=: n}), and note that T is a stopping time with respect to this
filtration. Furthermore lP'(T < oo) = 1 since Tis no larger than the firstpassage time to 0 of either
of the two singlecoordinate random walks, each of which has mean 0 and is therefore persistent.
Let a'f = var(X1) and ai = var(YI). We have that Un  Uo and Vn  Vo are sums of
independent summands with means 0 and variances a'f and ai respectively. It follows by considering
407
[12.9.18][12.9.19]
Solutions
Martingales
the martingales (Un Uo) 2 no} and (Vn Vo) 2 no} (see equation (12.5.14) and Exercise (10.2.2))
that
E{(UT  Uo) 2} = o}E(T), E{(VT  Vo) 2} = o}E(T).
Applying the same argument to (Un
E { (UT
+ VT Uo
Vo) 2} = E(T)E{(Xr
+ Yr) 2} =
E(T)(a12 + 2c + a 22).
+ cE(T 1\ m).
Therefore E(T 1\ m) :S E(Uo Vo)/(21cl) for all m, implying that E(T) = limm+oo E(T
and so E(T) = E(Uo Vo)/c as before.
1\
m) < oo,
18. Certainly 0 :S Xn :S 1, and in addition Xn is measurable with respect to the afield :Fn =
a(Rr, R2, ... , Rn). Also E(Rn+l I Rn) = Rn Rn/(52 n), whence E(Xn+l I :Fn) = Xn.
Therefore Xn is a martingale.
A strategy corresponds to a stopping time. If the player decides to call at the stopping time T, he
wins with (conditional) probability X T and therefore IP'(wins) = E(XT ), which equals E(Xo) (=
by the optional stopping theorem.
Here is a trivial solution to the problem. It may be seen that the chance of winning is the same
for a player who, after calling "Red Now", picks the card placed at the bottom of the pack rather than
that at the top. The bottom card is red with probability
irrespective of the strategy of the player.
i)
i,
19. (a) A sums of money in weekt is equivalentto a sums /(1 +a)t in weekO, since the latter sum may
be invested now to yield s in week t. If he sells in week t, his discounted costs are :L~= 1 cI (1 + a )n
and his discounted profit is Xt/(1 + a)t. He wishes to find a stopping time for which his mean
discounted gain is a maximum.
Now
so that tt(T) = E{ (1
+ a)T ZT} 
(c/a).
(b) The function h(y) = ay  fy IP'(Zn > y) dy is continuous and strictly increasing on [0, 100 ),
with h(O) = E(Zn) < 0 and h(y) + oo as y + oo. Therefore there exists a unique y (> 0) such
that h (y) = 0, and we choose y accordingly.
(c) Let :Fn = a(Zr, Z2, ... , Zn). We have that
00
E(max{Zn, y}) = y
00
408
+ a)ny
Solutions [12.9.20][12.9.21]
Problems
Let ~t(r) be the mean gain of following the strategy 'accept the first offer exceeding r  (cja)'.
The corresponding stopping timeT satisfies JP>(T = n) = G(r)n(l G(r)), and therefore
00
~t(r)
+ (cja)
L E{ (1 + a)T Zr l{T=n)}
n=O
00
= L(l
+ a)nG(r)n(l G(r))E(Z, I z,
n=O
1 +a
{ r(l  G(r))
1 +a G(r)
00
> r)
(1  G(y)) dy } .
Differentiate with care to find that the only value of r lying in the support of Z 1 such that ~t' (r) = 0
is the value r = y. Furthermore this value gives a maximum for ~t( r ). Therefore, amongst strategies
of the above sort, the best is that with r = y. Note that ~t(Y) = y(l +a)  (cja).
Consider now a general strategy with corresponding stopping time T, where JP>(T < oo) = 1. For
anypositiveintegerm, T /\misaboundedstoppingtime, whenceE(VTAm):::: E(Vo) = y(l+a). Now
IVT Am I :::: l::~o IVil, and l::~o EIVi I < oo. Therefore {Vr Am : m :=::: 0} is uniformly integrable.
Also Vr Am + Vr a.s. as m + oo, and it follows that E(Vr Am) + E(Vr ). We conclude that
Jt(T) = E(Vr) (cja) :::: y(l +a) (cja) = Jt(y). Therefore the strategy given above is optimal.
(d) In the special case, JP>(ZI > y) = (y  1) 2 for y ::::, 2, whence y = 10. The target price is
therefore 9, and the mean number of weeks before selling is G(y)/(1  G(y)) = 80.
20. Since G is convex on [0, oo) wherever it is finite, and since G(l) =land G'(l) < 1, there exists
a unique value of IJ (> 1) such that G(l]) = IJ. Furthermore, Yn = IJZn defines a martingale with
mean E(Yo) = IJ. Using the maximal inequality (12.6.6),
IJ) ::S
kl
I]
E(supZn) ::S
n
L  .1
k=I I] 
+ n + 1) 2 + Mn+I I Fn}
= (Mn + n) 2  2(Mn + n)E(Xn+I
::S (Mn + n) 2 + Mn,
E{ (Mn+I
 1) + Mn
+ E{ (Xn+I
1 if Mn > 1,
if Mn = 1.
var(Xn+l I :Fn) = { 0
409
 1) 2  Xn+I :Fn}
[12.9.22][12.9.24]
Solutions
Martingales
+ n) 2 + Mn}
+ t) I.'Fs)
= E(fos W(u)du
+ 1s+t W(u)du
HW(s
W(s)l 2
I.'Fs) = M(s)
as required. We apply the optional stopping theorem (12.7.12) with the stopping timeT= inf{ u :
W ( u) E {a, b}}. The hypotheses of the theorem follow easily from the boundedness of the process
fort E [0, T], and it follows that
3 ) = ~a 3 (~) + ~b 3 (a).
~E(W(T)
3
3
ab
3
ab
[We have used the optional stopping theorem twice actually, in that E(W(T)) = 0 and therefore
JP>(W(T) =a)= bj(a b).]
23. With .'Fs = a(W(u) : 0 _:::: u _:::: s), we have for s < t that
E(R(t) 2 I .'Fs) = E(IW(s)l 2 + IW(t) W(s)l 2
+ 2W(s) (W(t)
+ (t s),
and the first claim follows. We apply the optional stopping theorem (12.7.12) with T = inf{u :
IW(u)l =a}, as in Problem (12.9.22), to find that 0 = E(R(T) 2  T) = a 2  E(T).
24. We apply the optional stopping theorem to the martingale W(t) with the stopping timeT to find
that E(W(T)) = a(1  Pb) + bpb = 0, where Pb = JP>(W(T) = b). By Example (12.7.10),
W (t) 2  t is a martingale, and therefore, by the optional stopping theorem again,
whence E(T) = ab. For the final part, we take a = band apply the optional stopping theorem to the
2 t] to obtain
martingale exp[eW(t)
ie
on noting that the conditional distribution ofT given W (T) = b is the same as that given W (T) = b.
Therefore, E(ei 02 T)
410
ie 2.
13
Diffusion processes
which suggest a diffusion approximation with instantaneous mean a(t, x) = (A. p,)x and instantaneous variance b(t, x) =(A.+ p,)x.
2. The following method is not entirely rigorous (it is an argument of the following wellknown
type: it is valid when it works, and not otherwise). We have that
aM =
at
1oo e
oo
0Yaf

at
dy
0Y f
dy,
by using the forward equation and integrating by parts. Assume that a(t, y) = I:n an (t)yn, b(t, y) =
I:n f3n (t)yn. The required expression follows from the 'fact' that
3.
with boundary condition M(O, e) = l. The solution is M(t) = exp{ ~e (2m + e)t}.
4.
with boundary condition M(O, e) = l. The characteristics of the equation are given by
dt
1
de
e
2dM
e2M'
411
[13.3.5][13.3.10]
Solutions
Diffusion processes
5. Fix t > 0. Suppose we are given W1 (s), W2(s), W3(s), for 0:::;: s :::;: t. By Pythagoras's theorem,
R(t + u) 2 =Xi+ X~+ X~ where the Xi are independent N(Wi (t), u) variables. Using the result of
Exercise (5.7.7), the conditional distribution of R(t + u) 2 (and hence of R(t + u) also) depends only
on the value of the noncentrality parameter e = R(t) 2 of the relevant noncentral x2 distribution.
It follows that R satisfies the Markov property. This argument is valid for the ndimensional Bessel
process.
6. By the spherical symmetry of the process, the conditional distribution of R(s +a) given R(s) = x
is the same as that given W(s) = (x, 0, 0). Therefore, recalling the solution to Exercise (13.3.5),
i
{ (ux) 2 +v 2 +w 2 }
(u,v,w):
dudvdw
2a
2:rra 312 exp u2+v2+w2:sy2 (
)
y 1 2:n: l:n:
1
{
1p=O
312 exp =0 0=0 (2:rra)
=loy~ { exp (
(p
~ax)2) 
(p ;ax)2)} dp,
I:Ft} =
9. (a) With s < t, S(t) = S(s) exp{a(t s) + b(W(t) W(s)) }. Now W(t) W(s) is independent
of {W(u) : 0:::;: u :::;: s}, and the claim follows.
(b) S (t) is clearly integrable and adapted to the filtration !F = (:Ft) so that, for s < t,
E(S(t) I Fs) = S(s)E(exp{a(t s) + b(W(t) W(s))} I Fs) = S(s)exp{a(t s) + ~b 2 (t s)},
which equals S(s) if and only if a+ ~b 2 = 0. In this case, E(S(t)) = E(S(O)) = 1.
10. Either find the instantaneous mean and variance, and solve the forward equation, or argue directly
as follows. With s < t,
lP'(S(t):::;: y I S(s) = x) = lP'(bW(t):::;: at+ logy I bW(s) =as+ logx).
Now b(W(t) W(s)) is independent of W(s) and is distributed as N(O, b2(t s)), and we obtain on
differentiating with respect to y that
f (t, y 1 s, x) _
1
( (log(y/x)a(ts)) 2 )
exp ,
2
yy2:rrb2(t s)
2b (t s)
412
x,y > 0.
Solutions
[13.4.1][13.6.1]
Certainly X has continuous sample paths, and in addition EIX (t) I < oo. Also, if s < t,
as required, where we have used the fact that W(t) W(s) is N(O, t  s) and is independent of :Fs.
2.
Apply the optional stopping theorem to the martingale X of Exercise (13.4.1), with the stopping
time T, to obtain E(X(T)) = 1. Now W(T) = aT + b, and therefore E(elfrT+ilib) = l where
1/J = i ae + 2 . Solve to find that
1e
E(elfrT)
3.
j_:
2me 2 mx g(t, y I x) dx
where g(t, y I x) = (2nt) 2 exp{ (y  x  mt) 2 j(2t)}. The first two terms tend to 0 as t + oo,
regardless of the sign of m. As for the integral, make the substitution u = (x  y mt) I ..fi to obtain,
as t+ oo,

(d+y+mt)/../i
00
I 2
ezu
{ 21mle21mly
2me 2my   du +
..,fiir
0
ifm < 0,
ifm:::: 0.
IP'(W(t)
>xI Z, W(O) =
0)
= lim IP'(W(t)
w!0
>xI Z, W(O) =
w)
where Z = {no zeros in (0, t]}; the small missing step here may be filled by conditioning instead on
the event {W(E) = w, no zeros in (E, t]}, and taking the limit as E .j, 0. Now, if w > 0,
= w) =
00
IP'(Z I W(O) =
w)
= l 2
i:
+ w)} dy
f(t, y)dy
[13.6.2][13.6.5]
Solutions
Diffusion processes
by a consideration of the minimum value of Won (0, t]. It follows that the density function of W(t),
conditional on Z n {W(O) = w}, where w > 0, is
Jw
+ w)
'
X> 0.
X> 0.
w),0
2.
C=;),
0}
=a
0}
= s(l s),
if 0 ::::: s ::::: t ::::: 1. Therefore the Brownian bridge B satisfies, for 0 ::::: s ::::: t ::::: 1,
E(B(s)B(t)) = E{ B(s)E(B(t) I B(s))} = 1  t E(B(s) 2 ) = s(l t)
1s
as required. Certainly E(B(s)) = 0 for all s, by symmetry.
3. W is a zeromean Gaussian process on [0, 1] with continuous sample paths, and also W(O) =
W(l) = 0. Therefore W is a Brownian bridge if it has the same autocovariance function as the
Brownian bridge, that is, c(s, t) = min{s, t} st. For s < t,
cov(W(s), W(t)) = cov(W(s) sW(1), W(t) tW(l)) = s ts st
+ st =sst
oo
E~)
n=O
n=O
.
1
.
1
hm IW(u)l::::: hm {IW(n)l
+1
n+oo n
u+oo u
+ Mn}+ 0
5.
a.s.
In the notation of Exercise (13.6.4), we are asked to calculate the probability that W has no zeros
in the time interval between s /(1  s) and t /(1  t). By Theorem (13.4.8), this equals
1
2
1   cos
:rr
s(1y)
2
1
t(1s) =;cos
414
V~
Solutions
Stochastic calculus
[13.7.1][13.7.5]
00
E(Wt
/{X~:o:x}) dx =
Hence E(X~ 2 ) :::;: 4E(Wh. Now X~ 2 is monotone increasing inn, and W has continuous sample
paths. By monotone convergence,
2.
3.
w?,
n+oo
4.
w?
w? +it,
n+oo
n+oo
iw?.
Thus U is a stationary Gaussian Markov process, namely the OmsteinUhlenbeck process. [See
Example (9.6.10).]
5.
1
1
e.B(u+v)min{u,v}dudv
u=O v=O
415
[13.8.1][13.8.1]
Solutions
Diffusion processes
after prolonged integration. By the linearity of the definition of U, it is a Gaussian process. From the
calculation above, it has autocovariance function c(s, s +t) = (ef3(ts) ef3(t+sl)j(2(3). From this
we may calculate the instantaneous mean and variance, and thus we recognize an OmsteinUhlenbeck
process. See also Exercise (13.3.4) and Problem (13.12.4).
tl
n1 1
(tj+1 tj) 3
. 0 2
n1 1 (t)3
L.
2 n
+ 0
asn+ oo .
]= 0
]=
Therefore,
lim ( tWtn'>00
n1
= tWt
j=O
lor Ws ds.
t
t w? L [
+
The fact that the last two terms tend to 0 in mean square may be verified in the usual way. For example,
E(I~(Yj+1 Yj) 3
J=O
n ~E[(Vj+l=
J=O
n1
= 6
(tj+ 1 j=O
Vj) 6]
n1
fj ) 3 =
416
L U)
j=O n
3
+ 0
as n+ oo.
Solutions
Ito'sjormula
(c) It was shown in Exercise (13.7.3a) that f~ Ws dWs = i
[13.8.2][13.9.1]
u) du = s 2
(t2
  s) .
3.
~).
v~t (~2
6t
ll1/f1
111/Fi II for
The question permits us to use the integrating factor ef3t to give, formally,
ef3t Xt
on integrating by parts. This is the required result, and substitution verifies that it satisfies the given
equation.
6.
1/f II
By the argument before equation (13.8.9), /(cp(n)) ~ /(1/f) as n + oo. By Lemma (13.8.4),
II/ (cp(n)) 112 = llc/J(n) II, and the claim follows.
The first claim follows by the Levy characterization of a Wiener process (12. 7.1 0).
417
[13.9.2][13.10.2]
Diffusion processes
Solutions
We have inn dimensions that R2 = XI +X~++ X~, and the same argument yields that
(Xi/ R) dXi is a Wiener process. By Example (13.9.7) and the above,
Zt = ~i
Jd
d(R 2) = 2
i=l R
i=I
2.
Applying Ito's formula (13.9.4) to Y1 = W 14 we obtain dY1 = 4W? dW1 + 6W? dt. Hence,
3.
Apply Ito's formula (13.9.4) to obtain dY1 = W 1 dt + t dW1 Cf. Exercise (13.8.1).
4.
(a) (1
+ t) dX
(c)
+ dW,
+ v'l=X2 dW,
= X dt
(b) dX = ~X dt
1.
(aez K)
00
~ exp ( (z ;) 2 )
v2nr2
lrog(Kfa)
I 2
ezY
(aeY+<Y  K )   dy
= aeY+z'
21
a
00
I
2
ez(yr)
v2n
dz
2r
z y
log(K/a) y
where y =   , a=     r
dy K<l>(a)
(b) We have that ST = ae 2 where a = S1 and, under the relevant conditional Qdistribution, Z is
normal with mean y = (r  ~a 2 )(T  t) and variance r 2 = a 2(T  t). The claim now follows by
the result of part (a).
2. (a) Set ~(t, S) = ~(t, S1 ) and 1/r(t, S) = 1/r(t, S1 ), in the natural notation. By Theorem (13.10.15),
we have 1/rx = 1fr1 = 0, whence 1/r(t, x) = c for all t, x, and some constant c.
(b) Recall that dS = MS dt +aS dW. Now,
d(~S + 1frer 1 ) = d(S 2 + 1frer 1 ) = (aS) 2 dt + 2S dS + ert dlfr + lfrrert dt,
418
Solutions [13.10.3][13.10.5]
Option pricing
Jd Su du satisfies dZt
Zt dSt
s[ dt, whence
d(~ S + 1/fert)
3. We need to check equation (13.10.4) remembering that dMt = 0. Each of these portfolios is
selffinancing.
(a) This case is obvious.
(b)d(~S + 1/J)
4. The time of exercise of an American call option must be a stopping time for the filtration (:Ft).
The value of the option, if exercised at the stopping time r, is V, = (S,  K)+, and it follows by
the usual argument that the value at time 0 of the option exercised at r is E<Qi(err V, ). Thus the
value at time 0 of the American option is sup, {E<Qi(err V,) }, where the supremum is taken over all
stopping times r satisfying lP'( r :::= T) = 1. Under the probability measure Q, the process ert Vt is
a martingale, whence, by the optional stopping theorem, E<Qi(err V,) = Vo for all stopping times r.
The claim follows.
5. We rewrite the value at time 0 of the European call option, possibly with the aid of Exercise
(13.10.1), as
where N is an N(O, 1) random variable. It is immediate that this is increasing in So and r and is
decreasing in K. To show monotonicity in T, we argue as follows. Let T1 < T2 and consider the
European option with exercise date T2. In the corresponding American option we are allowed to
exercise the option at the earlier time T1 . By Exercise (13.10.4), it is never better to stop earlier than
T2, and the claim follows.
Monotonicity in a may be shown by differentiation.
419
[13.11.1][13.12.2]
Solutions
Diffusion processes
!!._
dr
lP'( G before H
II W(O)I = r ).
(rdldPR) =O
dr
since PR is spherically symmetric. Solve subject to the boundary equations PR(E) = 1, PR(R) = 0,
to obtain
2d  R2d
r
d2
as R+ oo.
PR(r) = 2d R2d + (E/r)
E
2. The electrical resistance Rn between 0 and the set /':,.n is no smaller than the resistance obtained
by, for every i = 1, 2, ... , 'shorting out' all vertices in the set /':,.i. This new network amounts to a
linear chain of resistances in series, points labelled /':,.i and /':,.i+ 1 being joined by a resistance if size
Nil . It follows that
00
1
R(G)= lim Rn::::L
n>oo
i=O Ni
By Theorem (13.11.18), the walk is persistent if "Ei Nil = oo.
3. Thinking of G as an electrical network, one may obtain the network H by replacing the resistance
of every edge e lying in G but not in H by oo. Let 0 be a vertex of H. By a well known fact in the
theory of electrical networks, R(H) :::: R(G), and the result follows by Theorem (13.11.19).
f3 =
_!!__ +s2
s+t
(! _1_)
s
s+t
= t.
2. Certainly W is Gaussian with continuous sample paths and zero means, and it is therefore sufficient
to prove that cov(W (s), W(t)) = min{s, t}. Now, if s .:::; t,
as required.
420
Solutions [13.12.3][13.12.4]
Problems
Certainly U is Gaussian with zero means, and U (0) = 0. Now, with s1 = e 2f3t  1,
E{U(t +h) I U(t)
+ h) 2
= s1 +
U(t)
It follows that
u}
4.
aM
at
aM
aB
1 2 2
={38+a B M
2
with boundary condition M (B, 0) = e 8 u. Solve this equation (as in the exercise given) to obtain
the moment generating function of the given normal distribution. Now M (t, B) + exp{ ~B 2 a 2 j (2{3)}
as t + oo, whence by the continuity theorem V (t) converges in distribution to the N (0, ~a 2 j fJ)
distribution.
If V (0) has this limit distribution, then so does V (t) for all t. Therefore the sequence (V Ct1), ... ,
V(tn)) has the same joint distribution as (V(t1 +h), ... , V(tn +h)) for all h, t1; ... , tn, whenever
V (0) has this normal distribution.
In the stationary case, E(V(t)) = 0 and, for s ::; t,
cov(V(s), V(t))
where c(O) = var(V (s)); we have used the first part here. This is the autocovariance function
of a stationary Gaussian Markov process (see Example (9.6.10)). Since all such processes have
autocovariance functions of this form (i.e., for some choice of fJ), all such processes are stationary
OmsteinUhlenbeck processes.
The autocorrelation function is p(s) = ef31sl, which is the characteristic function of the Cauchy
density function
1
X ER
f(x) = fJn{l + (x/{3)2},
421
[13.12.5][13.12.7]
5.
Diffusion processes
Solutions
aM
at
aM + {JB
1
2 aM
ae 2 ae '
= aB
dM
dt
+ ifJB))
E(D(t)) =
2d()
2aB
aM I =
ae 8=0
a2 M
E(D(t)2) = 2
+ f3() 2
ae 8=0
+ ifJB))
= e 8d.
deat,
fJd
= eat
(eat  1)
+ d2e2at'
lim
8>oo
2adeat
{3(1  ea )
6.
The equilibrium density function g(y) satisfies the (reduced) forward equation
d
(ag)
dy
where a(y)
are
1 d
+ (bg)
=
2 dy2
y = c,d.
Integrate (*) from c to y, using the boundary conditions, to obtain
c
~d.
Integrate again to obtain g(y) = Aef3Y 2fa 2 . The constant A is given by the fact that f~c g(y) dy = 1.
7.
First we show that the series converges uniformly (along a subsequence), implying that the limit
exists and is a continuous function of t. Set
n1
. (k )
~ sm t
Zmn(t) = ~ kXk>
Mmn = sup{IZmn(t)l: 0 ~ t ~
7r }.
k=m
We have that
M2 <
mn 
su
n1 eikt
~ X
P l~ k k
O::or:::,:rr k=m
12
n1 x2
< ~ ___k_
~ k2
k=m
422
+2
nm11n11 XX I
1 J+l
~
~
~
~ ( + l)
1=1
J=m 1 1
Solutions
Problems
[13.12.8][13.12.8]
The mean value of the final term is, by the CauchySchwarz inequality, no larger than
nm1
nl1
L:
"
+ [)2
L.J j2(j
}=m
1=1
< 2(n  m)

F.t
m
m4

It follows that
implying that .2:::~ 1 M2nI 2n < oo a.s. Therefore the series which defines W converges uniformly
with probability 1 (along a ;ubsequence), and hence W has (a.s.) continuous sample paths.
Certainly W is a Gaussian process since W(t) is the sum of normal variables (see Problem
(7.11.19)). Furthermore E(W(t)) = 0, and
st
2 ~ sin(ks) sin(kt)
+ L.J
T(
T(
k=1
since the Xi are independent with zero means and unit variances. It is an exercise in Fourier analysis
to deduce that cov(W(s), W(t)) = min{s, t}.
8.
IYI <
b,
if IYI :S b,
g(t, y) =
0 if IYI =b.
Let g(t, y I d) be the N(d, t) density function, and note that g(, I d) satisfies (*) for any
'source' d. Let
00
L:
g(t, y) =
c1)kg(t, y 12kb),
k=00
a series which converges absolutely and is differentiable term by term. Since each summand satisfies
(*),so does the sum. Now g(O, y) is a combination of Dirac delta functions, one at each multiple of2b.
Only one such multiple lies in [b, b], and hence g(y, 0) = 8ao Also, setting y = b, the contributions
from the sources at 2(k 1)b and 2kb cancel, so that g(t, b) = 0. Similarly g(t, b) = 0, and
therefore g is the required solution.
Here is an alternative method. Look for the solution to(*) of the form el.nt sin{inn(y + b)jb};
such a sine function vanishes when IYI =b. Substitute into(*) to obtain An = n 2 n 2 /(8b 2 ). A linear
combination of such functions has the form
00
g(t,y) =Lane
Ant .
sm
n=1
423
nn(y +b)
) .
2b
[13.12.9][13.12.11]
Solutions
Diffusion processes
We choose the constants an such that g(O, y) = 8yo for IYI < b. With the aid of a little Fourier
analysis, one finds that an = b 1 sin(~mr).
Finally, the required probability equals the probability that wa has been absorbed by time t, a
probability expressible as 1 f~b fa(t, y) dy. Using the second expression for
this yields
r,
9. Recall that U(t) = e 2mD(t) is a martingale. LetT be the time of absorption, and assume that
the conditions of the optional stopping theorem are satisfied. Then E(U(O)) = E(U(T)), which is to
say that 1 = e2ma Pa + e2mb(l  Pa).
= lP'(W(t)
o),
giving that
apr (b)
az;=
where f(t, x) is the N(O, t) density function. Now, using conditional probabilities,
lP'(F(O, t) I W(O) =a, W(t) =b)=
1
apr(b) = 1 _ e2abft.
f(t, b a) ab
if 0 < s < t. The claim follows since F (to, t2) s; F (t1, t2).
(c) Remember that sinx = x + o(x) as x 0. Take the limit in part (b) as to
+0 to obtain ,fiJ!i2.
11. Let M(t) = sup{W(s): 0:::; s:::; t} and recall that M(t) has the same distribution as IW(t)l. By
symmetry,
lP'( sup IW(s)l::::
w) :::; 2lP'(M(t):::: w)
= 2lP'(IW(t)l::::
w).
O~s~t
By Chebyshov's inequality,
424
Solutions [13.12.12][13.12.13]
Problems
Therefore L:n lP'(An(E)) < oo, implying by the BorelCantelli lemma that (a.s.) only finitely many of
the An(E) occur. Therefore t 1W(t)+ 0 a.s. as t+ oo. Compare with the solution to the relevant
part of Exercise (13.6.4).
v2 p
0 ifwEH,
{ 1 'f
G.
1 W E
p(r, e) =
Certainly combinations having this form satisfy Laplace's equation, and the boundary condition gives
that
00
H(e) = bo
+L
{an sin(ne)
+ bn cos(ne) },
1e1
< n,
n=1
where
H(e) = {
if  n <
e < o,
ifO<e<n.
The collection {sin(me), cos( me) : m 2: 0} are orthogonal over ( n, n). Multiply through(*) by
sin(me) and integrate over ( n, n) to obtain nam = {1  cos(nm)}/m, and similarly bo = ~ and
bm = 0 for m 2: 1.
13. The joint density function of two independent N (0, t) random variables is (2n t)  1 exp{ (x 2 +
y2)j(2t)}. Since this function is unchanged by rotations of the plane, it follows that the two coordinates of the particle's position are independent Wiener processes, regardless of the orientation of the
coordinate system. We may thus assume that lis the line x = d for some fixed positive d.
The particle is bound to visit the line l sooner or later, since lP'(W1 (t) < d for all t) = 0. The
firstpassage timeT has density function
t > 0.
Conditional on {T = t}, D = W2(T) is N(O, t). Therefore the density function of Dis
fv(u) =
loo
oo
fDIT(u I t)fT(t) dt =
looo e(u
d
+d )/( 2
2
o 2m 2
t) dt
d
n(u 2
425
2 + ;
1e1
< ~n.
+ d2)
u E ffi:.,
[13.12.14][13.12.18]
Solutions
Diffusion processes
14. By an extension of Ito's formula to functions of two Wiener processes, U = u(Wr, W2 ) and
V = v(Wr, W2) satisfy
dU = Ux dWr
dV = Vx dWr
where Ux, Vyy, etc, denote partial derivatives of u and v. Since is analytic, u and v satisfy the
CauchyRiemann equations Ux = vy, uy = vx, whence u and v are harmonic in that Uxx + uyy =
Vxx + Vyy = 0. Therefore,
The matrix (
~xUy
uy) is an orthogonal rotation oflH:2 when u~ +u~ = 1. Since the joint distribution
Ux
of the pair (Wr, W2) is invariant under such rotations, the claim follows.
15. One method of solution uses the fact that the reversed Wiener process {W (t s) W (t) : 0 ::: s :::
t} has the same distribution as {W(s) : 0 ::: s ::: t}. Thus M(t) W (t) = maxo~s~dW(s) W(t)} has
the same distribution as maxo~u~dW(u) W(O)} = M(t). Alternatively, by the reflection principle,
lP'(M(t) 2':: x, W(t)::: y) = lP'(W(t) 2':: 2x y)
By differentiation, the pair M(t), W(t) has joint density function 2' (2x  y) for y ::: x, x 2':: 0,
where is the density function of the N(O, t) distribution. Hence M(t) and M(t) W(t) have the
joint density function 2'(x + y). Since this function is symmetric in its arguments, M(t) and
M(t) W(t) have the same marginal distribution.
16. The Lebesgue measure A(Z) is given by
A(Z)
= fooo lP'(W(t) = u) dt = 0.
= maxx 9 ~y W(s).
Then
+ {W(c) W(b)}
a~s~b
Since the three terms on the right are independent and continuous random variables, it follows that
lP'((M(c, d) = M(a, b)) = 0. Since there are only countably many rationals, we deduce that
lP'( (M(c, d) = M(a, b) for all rationals a < b < c <d) = 1, and the result follows.
18. The result is easily seen by exhaustion to be true when n = 1. Suppose it is true for all m ::: n  1
where n 2':: 2.
(i) If sn ::: 0, then (whatever the final term of the permutation) the number of positive partial sums and
the position of the first maximum depend only on the remaining n  1 terms. Equality follows by the
induction hypothesis.
(ii) If Sn > 0, then
n
Ar =
L Ar1 (k),
k=l
426
Solutions
Problems
[13.12.19][13.12.20]
where Ar1 (k) is the number of permutations with Xk in the final place, for which exactly r  1 of the
first n 1 terms are strictly positive. Consider a permutation rc = (xi 1 , xi 2 , ... , XinI, Xk) with Xk in
the final place, and move the positionofxk to obtain thenewpermutationrc' = (xk. xi 1 , Xi 2 , . , Xinl ).
The first appearance of the maximum in rc' is at its rth place if and only if the first maximum of the
reduced permutation (xi 1 , Xi 2 , ... , X in I) is at its (r  l)th place. [Note that r = 0 is impossible
since sn > 0.] It follows that
n
Br =
Br1 (k),
k=1
where Br1 (k) is the number of permutations with Xk in the final place, for which the first appearance
of the maximum is at the (r  1)th place.
By the induction hypothesis, Ar1 (k) = Br1 (k), since these quantities depend on the n  1
terms excluding Xk. The result follows.
2::f=
= W(jtjn) W((j
l)t/n), so that Sm
= W(mtjn).
Thus An
m;:tx W(jtjn) }.
O:'OJ:'On
By Problem (13.12.17), Rn ~ R as n+ oo. By Problem (13.12.16), the time spent by W at zero
is a.s. a null set, whence An ~ A. Hence A and R have the same distribution. We argue as follows
to obtain that that L and R have the same distribution. Making repeated use of Theorem (13.4.6) and
the symmetry of W,
lP'(L < x) =
lP'c~~~t W(s)
<
= 2lP'( sup {W(s) W(x)} < W(x)) = 2lP'(IW(t) W(x)l < W(x))
x:'Os:'Ot
= lP'(
= lP'(R
::": x).
0:'0S:'OX
Finally, by Problem (13.12.15) and the circular symmetry of the joint density distribution of two
independent N(O, 1) variables U, V,
lP'(IW(t) W(x)l < IW(x)l) = lP'((t x)V 2 ::::: xU 2 ) = lP' (
20. Let
Tx = {
= x}
:::::
~)
= ~ sin 1 ~.
t
Vt
427
u2 +V 2
Jr
[13.12.21][13.12.24]
Diffusion processes
Solutions
and similarly Vx = sup{t .::; 1 : W(t) = x}, with Vx = 1 if W(t) "# x for all t E [0, 1]. Recall that
Uo and Vo have an arc sine distribution as in Problem (13.12.19). On the event {Ux < 1}, we may
write (using the rescaling property of W)
Ux
Tx
+ (1
 Tx)Uo,
Vx
Tx
+ (1 
Tx)Vo,
where Uo and Yo are independent of Ux and Vx, and have the above arc sine distribution. Hence Ux
and Vx have the same distribution. Now Tx has the first passage distribution of Theorem (13.4.5),
whence
 (r r/J)
Tx,Uo
'
x2)} {
1
}
n./r/J(1 rp)
Therefore,
and
1
fux(u)= lo u frxux(t,u)du=
exp ( x2)
 ,
o
'
n.ju(1u)
2u
O<x<l.
21. Note that Vis a martingale, by Theorem (13.8.11). Fix t and let 1/fs = sign(Ws), 0.::; s .::; t.
We have that 111/fll = v'f, implying by Exercise (13.8.6) that E(V?) = ll/(1/f)ll~ = t. By a similar
calculation, E(V? 1 :Fs) = v? + t  s for 0 .::; s .::; t. That is to say, v?  t defines a martingale, and
the result follows by the Levy characterization theorem of Example (12.7.10).
22. The mean cost per unit time is
f.l(T) =
23. Consider the portfolio with Ht, S1) units of stock and 1/f(t, S1) units of bond, having total value
w(t, S1) = x~(t. x) + ert 1/f(t, S1). By assumption,
(1 y)x~(t, x) = yer 1 1jf(t, x).
Differentiate this equation with respect to x and substitute from equation (13.10.16) to obtain the
differential equation (1 y)~ + x~x = 0, with solution ~(t, x) = h(t)xYI, for some function h(t).
We substitute this, together with(*), into equation (13.10.17) to obtain that
h' h(l y)(~yo 2
+ r) =
0.
It follows that h(t) = A exp{(1  y)( ~yo 2 + r)t}, where A is an absolute constant to be determined
according to the size of the initial investment. Finally, w(t, x) = y 1 x~(t. x) = y 1h(t)xY.
24. Using Ito's formula (13.9.4), the drift term in the SDE for U1 is
(ut(T t, W)
where u1 and u22 denote partial derivatives of u. The drift function is identically zero if and only if

U[ 
zU22.
428
Bibliography
Samuel Johnson
429
Index
Abbreviations used in this index: c.f. characteristic function; distn distribution; eqn equation;
fn function; m.g.f. moment generating function; p.g.f. probability generating function; pr.
process; r.v. random variable; r.w. random walk; s.r.w. simple random walk; thm theorem.
A
absolute value of s.r. w. 6.1.3
absorbing barriers: s.r.w. 1.7.3,
3.9.1, 5, 3.11.23, 2526,
12.5.45, 7; Wiener pr.
12.9.223, 13.12.89
absorbing state 6.2.1
adapted process 13.8.6
affine transformation 4.13.11;
4.14.60
agedependent branching pr.
5.5.12; conditional 5.1.2;
honest martingale 12.9.2; mean
10.6.13
age, see current life
airlines 1.8.39, 2. 7. 7
alarm clock 6.15.21
algorithm 3.11.33, 4.14.63, 6.14.2
aliasing method 4.11.6
alternating renewal pr. 10.5.2,
10.6.14
American call option 13.10.4
analytic fn 13.12.14
ancestors in common 5.4.2
anomalous numbers 3.6.7
Anscombe's theorem 7.11.28
antithetic variable 4.11.11
ants 6.15.41
arbitrage 3.3.7, 6.6.3
Arbuthnot, J. 3.11.22
arc sine distn 4.11.13; sample
from 4.11.13
arc sine law density 4.1.1
arc sine laws for r.w.: maxima
3.11.28; sojourns 5.3.5; visits
3.10.3
B
babies 5.10.2
backward martingale 12.7.3
bagged balls 7.11.27, 12.9.1314
balance equations 11.7.13
Bandrika 1.8.3536, 4.2.3
bankruptcy, see gambler's ruin
Barker's algorithm 6.14.2
barriers: absorbing/retaining in
r.w. 1.7.3, 3.9.12, 3.11.23,
2526; hitting by Wiener pr.
13.4.2
Bartlett: equation 13.3.24;
theorem 8.7.6, 11.7.1
batch service 11.8.4
baulking 8.4.4, 11.8.2, 19
Bayes's formula 1.8.14, 1.8.36
bears 6.13.1, 10.6.19
Benford's distn 3.6.7
Berkson's fallacy 3.11.37
Bernoulli: Daniel 3.3.4, 3.4.34;
Nicholas 3.3.4
430
Index
7.11.19; c.f. 5.8.11; positive
c
cake, hot 3.11.32
call option: American 13.10.4;
European 13.10.45
CampbellHardy theorem 6.13.2
capturerecapture 3.5.4
car, parking 4.14.30
cards 1.7.2, 5, 1.8.33
Carroll, Lewis 1.4.4
casino 3.9.6, 7.7.4, 12.9.16
Cauchy convergence 7.3.1; in
m.s. 7.11.11
Cauchy distn 4.4.4; maximum
7 .11.14; moments 4.4.4;
431
D
dam 6.4.3
dead period 10.6.67
deathimmigration pr. 6.11.3
decay 5.12.48, 6.4.8
decimal expansion 3.1.4, 7 .11.4
decomposition: Cholesky 4.14.62;
Krickeberg 12.9.11
degrees of freedom 5.7.78
delayed renewal pr. 10.6.12
de Moivre: martingale 12.1.4,
12.4.6; trial 3.5.1
De Morgan laws 1.2.1
density: arc sine 4.11.13; arc
sine law 4.1.1; betsa 4.11.4,
Index
432
E
Eddington's controversy 1.8.27
editors 6.4.1
eggs 5.12.13
Ehrenfest model 6.5.5, 6.15.36
eigenvector 6.6.12, 6.15.7
embarrassment 2.2.1
empires 6.15.10
empirical distn 9.7.22
entrance fee 3.3.4
entropy 7 .5.1; conditional
6.15.45; mutual 3.6.5
epidemic 6.15.32
equilibrium, see stationary
equivalence class 7 .1.1
ergodic: coefficient 6.14.4;
measure 9.7.11; stationary
measure 9.7.11
ergodic theorem: Markov chain
6.15.44, 7.11.32; Markov pr.
7.11.33, 10.5.1; stationary pr.
9.7.1011, 13
Erlang's loss formula 11.8.19
error 3.7.9; of prediction 9.2.2
estimation 2.2.3, 4.5.3, 4.14.9,
7.11.31
Euler: constant 5.12.27, 6.15.32;
product 5.12.34
European call option 13.10.45
event: of convergence 7.2.6;
exchangeable 7.3.45; invariant
9.5.1; sequence 1.8.16; tail
7.3.3
excess life 10.5.4; conditional
10.3.4; and current 10.6.9;
limit 10.3.3; Markov 8.3.2,
10.3.2; moments 10.3.3;
Poisson 6.8.3, 10.3.1, 10.6.9;
reversed 8.3.2; stationary
10.3.3
exchangeability 7.3.45
expectation: conditional
3.7.23, 4.6.2, 4.14.12,
7.9.4; independent r.v.s 7.2.3;
linearity 5.6.2; tail integral
4.3.3, 5; tail sum 3.11.13,
4.14.3
exponential distn: c.f. 5.8.8;
holding time 11.2.2; in Poisson
pr. 6.8.3; lackofmemory
Index
property 4.14.5; limit in
branching pr. 5.6.2, 5.12.21,
6. 7.1; limit of geometric distn
5.12.39; heavy traffic limit
11.6.1; distn of maximum
4.14.18; in Markov pr. 6.8.3,
6.9.9; order statistics 4.14.33;
sample from 4.14.48; sum
4.8.1, 4, 4.14.10, 5.12.50,
6.15.42
exponential martingale 13.3.9
exponential smoothing 9.7.2
extinction: of birthdeath pr.
6.11.3, 6.15.25, 27, 12.9.10;
of branching pr. 6.7.23
extremevalue distn 4.1.1,
4.14.46, 5.12.34, 7.11.13; c.f.
and mean 5.12.27
F
F(r, s) distn 4.10.2, 4;
noncentral 5.7.8
fair fee 3.3.4
families 1.5.7, 3.7.8
family, planning 3.11.30
Farkas's theorem 6.6.2
filter 9. 7.2
filtration 12.4.12, 7
fingerprinting 3.11.21
finite: Markov chain 6.5.8, 6.6.5,
6.15.4344; stopping time
12.4.5; waiting room 11.8.1
first passage: c.f. 5.10.78;
diffusion pr. 13.4.2; distn
5.10.78, 5.12.1819; Markov
chain 6.2.1, 6.3.6; Markov pr.
6.9.56; mean 6.3.7; m.g.f.
5.12.18; s.r.w. 5.3.8; Wiener
pr. 13.4.2
first visit by s.r.w. 3.10.1, 3
Fisher: spherical distn 4.14.36;
F.TippettGumbel distn
4.14.46
FKG inequality 3.11.18, 4.11.11
flipflop 8.2.1
forest 6.15.30
Fourier: inversion thm 5.9.5;
series 9.7.15, 13.12.7
fourth moment strong law 7 .11.6
fractional moments 3.3.5, 4.3.1
function, upperclass 7.6.1
functional eqn 4.14.5, 19
H
HajekRenyiChow inequality
12.9.45
Hall, Monty 1.4.5
Hastings algorithm 6.14.2
Hawaii 2.7.17
hazard rate 4.1.4, 4.4.7; technique
4.11.10
heat eqn 13.12.24
Heathrow 10.2.1
heavy traffic 11.6.1, 11.7.16
hen, see eggs
HewittSavage zeroone law
7.3.45
hitting time 6.9.56; theorem
3.10.1, 5.3.8
Hoeffding's inequality 12.2.12
Holder's inequality 4.14.27
holding time 11.2.2
homogeneous Markov chain 6.1.1
honest birthdeath pr. 6.15.26
Hotelling's theorem 4.14.59
house 4.2.1, 6.15.20, 51
hypergeometric distn 3.11.1011
hypoexponential distn 4.8.4
433
Index
434
Index
stationary 9.7.1112; strongly
mixing 9.7.12
median 2.7.11, 4.3.4, 7.3.11
menages 1.8.23
Mercator projection 6.13.5
meteorites 9. 7.4
metric 2.7.13, 7.1.4; Levy 2.7.13,
7 .1.4, 7 .2.4; total variation
2.7.13
m.g.f. inequality 5.8.2, 12.9.7
migration pr., open 11.7.1, 5
millionaires 3.9.4
Mills's ratio 4.4.8, 4.14.1
minimal, solution 6.3.67, 6.9.6
Minkowski's inequality 4.14.27
misprints 6.4.1
mixing, strong 9. 7.12
mixture 2.1.4, 2.3.4, 4.1.3, 5.1.9
modified renewall0.6.12
moments: branching pr. 5.4.1;
fractional 3.3.5, 4.3.1;
generating fn 5.1.8, 5.8.2,
5.11.3; joint 5.12.30; problem
5.12.43; renewal pr. 10.1.1; tail
integral 4.3.3, 5.11.3
Monte Carlo 4.14.9
Monty Hall 1.4.5
Moscow 11.8.3
moving average 8.7.1, 9.1.3,
9.4.2, 9.5.3, 9.7.12, 7;
spectral density 9. 7. 7
multinomial distn 3.5.1; marginals
3.6.2; p.g.f. 5.1.5
multinormal distn 4.9.2; c.f.
5.8.6; conditioned 4.9.67;
covariance matrix 4.9.2;
maximum 5.9.7; sampling
from 4.14.62; standard 4.9.2;
transformed 4.9.3, 4.14.62
Murphy's law 1.3.2
mutual information 3.6.5
N
needle, Buffon's 4.5.2,
4.14.3132
negative binomial distn 3.8.4;
bivariate 5.12.16; moments
5.12.4; p.g.f. 5.1.1, 5.12.4
negative hypergeometric distn
3.5.4
Newton, I. 3.8.5
noncentral distn 5.7.78
nonhomogeneous: birth pr.
6.15.24; Poisson pr. 6.13.7,
6.15.1920
noodle, Buffon's 4.14.31
0
occupation time for Wiener pr.
13.12.20
open migration 11.7.1, 5
optimal: packing 12.2.1; price
12.9.19; reset time 13.12.22;
serving 11.8.13
optimal stopping: dice 3.3.89;
marriage 4.14.35
optional stopping 12.5.18,
12.9.19; diffusion 13.4.2;
Poisson 12.7.2
order statistics 4.14.21;
exponential 4.14.33; general
4.14.21; marginals 4.14.22;
uniform 4.14.2324, 39,
6.15.42, 12.7.3
OmsteinUhlenbeck pr. 9.7.19,
13.3.4, 13.7.45, 13.12.34, 6;
reflected 13.12.6
orthogonal: increments 7.7.1;
polynomials 4.14.37
osmosis 6.15.36
p
pairwise independent: events
1.5.2; r.v.s 3.2.1, 3.3.3, 5.1.7
paradox: Bertrand 4.14.8;
Borel 4.6.1; Carroll 1.4.4;
Galton 1.5.8; inspection
10.6.5; Parrando 6.15.48; St
Petersburg 3.3.4; voter 3.6.6
parallel lines 4.14.52
parallelogram 4.14.60; property
7.1.2;
435
parking 4.14.30
Parrando's paradox 6.15.48
particle 6.15.33
partition of sample space 1.8.10
Pasta property 6.9.4
patterns 1.3.2, 5.2.6, 5.12.2,
10.6.17, 12.9.16
pawns 2.7.18
Pepys's problem 3.8.5
periodic state 6.5.4, 6.15.3
persistent: chain 6.4.10, 6.15.6;
r.w. 5.12.56, 6.3.2, 6.9.8,
7.3.3, 13.11.2; state 6.2.34,
6.9.7
Petersburg, see St Petersburg
pig 1.8.22
points, problem of 3.9.4, 3.11.24
Poisson: approximation 3.11.35;
coupling 4.12.2; flips 3.5.2,
5.12.37; sampling 6.9.4
Poisson distn 3.5.3;
characterization of 5.12.8, 15;
compound 3.8.6, 5.12.13; and
gamma distn 4.14.11; limit of
binomial 5.12.39; modified
3.1.1; sum 3.11.6, 7.2.10
Poisson pr. 6.15.29; age 10.5.4;
arrivals 8.7.4, 10.6.8;
autocovariance 7 .11.5,
9 .6.1; characterization
6.15.29, 9.7.16; colouring
theorem 6.15.39; compound
6.15.21; conditional property
6.13.6; continuity in m.s.
7 .11.5; covariance 7 .11.5;
differentiability 7.11.5;
doubly stochastic 6.15.2223;
excess life 6.8.3, 10.3.1;
forest 6.15.30; gaps 10.1.2;
Markov renewal pr. 8.3.5,
10.6.9; martingales 12.7.2;
nonhomogeneous 6.13.7,
6.15.1920; optional stopping
12.7.2; perturbed 6.15.3940;
renewal 8.3.5, 10.6.910;
Renyi's theorem 6.15.39;
repairs 11.7.18; sampling
6.9.4; spatial 6.15.3031,
7.4.3; spectral density
9.7.6; sphere 6.13.34;
stationary increments 9.7.6,
16; superposed 6.8.1; thinned
6.8.2; total life 10.6.5; traffic
6.15.40, 49, 8.4.3
poker 1.8.33; dice 1.8.34
P6lya's urn 12.9.1314
portfolio 13.12.23; selffinancing
13.10.23
positive definite 9.6.1, 4.9.1
Index
positive state, see nonnull
postage stamp lemma 6.3.9
potential theory 13.11.13
power series approximation
7.11.17
Pratt's lemma 7.10.5
predictable step fn 13.8.4
predictor: best 7.9.1; linear 7.9.3,
9.2.12, 9.7.1, 3
probabilistic method 1.8.28, 3.4. 7
probability: continuity 1.8.16,
1.8.18; p.g.f. 5.12.4, 13; vector
4.11.6
problem: matching 3.4.9, 3.11.17,
5.2.7, 12.9.21; menages 1.8.23;
Pepys 3.8.5; of points 3.9.4,
3.11.24; Waldegrave 5.12.10
program, dual, linear, and primal
6.6.3
projected r.w. 5.12.6
projection theorem 9.2.10
proofreading 6.4.1
proportion, see empirical ratio
proportional investor 13.12.23
prosecutor's fallacy 1.4.6
protocol 1.4.5, 1.8.26
pullthrough property 3.7.1
Q
quadratic variation 8.5.4, 13.7.2
queue: batch 11.8.4; baulking
8.4.4, 11.8.2, 19; busy period
6.12.1, 11.3.23, 11.5.1,
11.8.5, 9; costs 11.8.13;
departure pr. 11.2.7, 11.7.24,
11.8.12; difficult customer
11.7.4; D/M/1 11.4.3, 11.8.15;
dual 11.5.2; Erlang's loss fn
11.8.19; finite waiting room
11.8.1; G/G/1 11.5.1, 11.8.8;
G/M/1 11.4.12, 11.5.23;
heavy traffic 11.6.1, 11.8.15;
idle period 11.5.2, 11.8.9;
imbedded branching 11.3.2,
11.8.5, 11; imbedded Markov
pr. 11.2.6, 11.4.1, 3, 11.4.1;
imbedded renewal 11.3.3,
11.5.1; imbedded r.w. 11.2.2,
5; Markov, see M/M/1; M/D/1
11.3.1, 11.8.1011; M/G/1
11.3.3, 11.8.67; M/G/oo
6.12.4, 11.8.9; migration
system 11.7.1, 5; M/M/1 6.9.3,
6.12.1, 11.2.23, 56, 11.3.2,
11.6.1, 11.8.5, 12; M!Mik
11.7.2, 11.8.13; M/M/oo 8.7.4;
series 11.8.3, 12; supermarket
11.8.3; tandem 11.2.7, 11.8.3;
taxicabs 11.8.16; telephone
436
Index
reliability 3.4.56, 3.11.1820,
renewal: age, see current life;
alternating 10.5.2, 10.6.14;
asymptotics 10.6.11; Bernoulli
8.7.3; central limit theorem
10.6.3; counters 10.6.68,
15; current life 10.3.2,
10.5.4, 10.6.4; delayed
10.6.12; excess life 8.3.2,
10.3.14, 10.5.4; r. function
10.6.11; gaps 10.1.2; key r.
theorem 10.3.3, 5, 10.6.11;
Markov 8.3.5; rn.g.f. 10.1.1;
moments 10.1.1; Poisson
8.3.5, 10.6.910; r. process
8.3.4; r.reward 10.5.14;
r. sequence 6.15.8, 8.3.1, 3;
stationary 10.6.18; stopping
time 12.4.2; sum/superposed
10.6.10; thinning 10.6.16
Renyi's theorem 6.15.39
repairman 11.7.18
repulsion 1.8.29
reservoir 6.4.3
resources 6.15.47
retaining barrier 11.2.4
reversible: birthdeath pr.
6.15.16; chain 6.14.1; Markov
pr. 6.15.16, 38; queue
11.7.23, 11.8.12, 14; r.w.
6.5.1
RiemarmLebesgue lemma 5.7.6
robots 3.7.7
rods 4.14.2526,
ruin 11.8.18, 12.9.12; see also
gambler's ruin
runs 1.8.21, 3.4.1, 3.7.10, 5.12.3,
4647
s
afield 1.2.2, 4, 1.8.3, 9.5.1,
9.7.13; increasing sequence
of 12.4.7
StJohn's College 4.14.51
St Petersburg paradox 3.3.4
sample: normal 4.10.5, 5.12.42;
ordered 4.12.21
sampling 3.11.36; Poisson 6.9.4;
with and without replacement
3.11.10
sampling from distn: arc sine
4.11.13; beta 4.11.45; Cauchy
4.11.9; exponential4.14.48;
gamma 4.11.3; geometric
4.11.8; Markov chain 6.14.3;
multinormal 4.14.62; normal
4.11.7, 4.14.49; s.r.w. 4.11.6;
uniform 4 .11.1
437
Index
438
Гораздо больше, чем просто документы.
Откройте для себя все, что может предложить Scribd, включая книги и аудиокниги от крупных издательств.
Отменить можно в любой момент.