Академический Документы
Профессиональный Документы
Культура Документы
1 and sets As,
S(m; Ai An)
SPA) = FE Pla) ee Play Ana)
An are n sets then
(-1)" (PULTAR) = Som; Ans es An)) 2 0
‘Use mathematical induction, The claim holds for m = 1 (by a)) (and also for m
by b)). Say the claim holds for m. Now ehow that this implies the claim holds for m +1, ie.,
that:
Subelai
(m is fixed) For each m > 1: If At,.++,Am are m sets then
(ay (PUTA) ~ S(m 415 An, An) 2 0.
Proof of subelaim: Again use mathematical induction. This subclaim holds for n = 1,....m
because the inclusion-exclusion formula (Exercise 12) shows that the left side equals zero. Sup-
ppose the claim holds for n. Now show that this implies the subelaim holds for n+ 1, ive.,
that
(Cy (PEUT# Ad — Sm + 154s
‘To do this, fist observe that
inta)) 20. (Oo)
Som 41, Atyoos An) # PlAnas) = Slo; Ay Ants sees AnAnss)
= S(m 41d Ansa):
‘Therefore
(1! (P(UZt Aa) — Sn 415 Ans... Antt))
(=1)™4" (PUPAL) + PlAnsi) = P(UPAAnds) — Sm 415 Atseos Ana)
(by inclusion exclusion for 2 sets)
= (0 (PUTA)
= PUPA Angi) + Sm Arne
(by identity (64))
my" (PUPAL) = S(m $1; Ai. An)
#(=1)" (PUUTALA mgt) = Sm ArAnsiyees Annes)
‘The first term is nonnegative by the induction hypothesis on n; the second term is nonnegative
by the induction hypothesis on m. So (+) holds, and we'te done!
vAnAngs) = Stim +1 Atye-- An)Section 1.4
Section 1.4
1. Let x denote the proportion of women in the population. Then the proportion r of righthanders is
given by 1 = 92x x +.88 x (I~ x) = 88+ 04x.
» 6
b) (i) :since 0 < <1 implies .88 P(lZnl < /n/33) 20.99 <> /n/33 > 2.58 <> n> 220.
30. D} takes the values 0, 1, 4, 9, 16, 25, 36, 49, 64, 81 with equal probability, so the X; are independent
‘with common distribution
z ofilals|els
PR t[2}2[4|2[2
So E(X:) = 9/2 and Var(X:) = 9.05 => SD(X:) = 3.008.
a) By the law of averages, you expect X,, to be close to E(Xx)
b) P(IXn — 4.5] > €)
= P (Sigg! > 2s) ~ P (I21> 2s) = 2” (2 > a)»
where Z has standard normal distribution. For n = 10000, we need ¢ such that
P[z> |
Aherefore ¢ = 2.81 x 3.008/100 = 0.085.
c) Need n such that P(|Xn— 4.5] < 0.01) > 0.99 i.e.,
4.5 for large n. So predict 4.5.
0025 + 9M = 2.81,
P [121 < 4x21] 2020 gS 2258
therefore w > 602276,
a) We have calculated (Xi) = 9/2 and Var(Xi) = 9.05. From the previous problem, we have
B(D:) = 9/2 and Var(Di) = 33/4 = 8.35. Since D; has smaller variance than does Xi, the
value of Dy can be predicted more accurately.
«) Since E(Xigo) = 4.5, you should predict the first digit of Xo to be 4. ‘The chance of being
correct is
Pla < Rio <5 w P (|ERgcen|
<
&
per> e+)
©) BW) = B(P 1) = £(r)
Var(T = 1) = Var(T) = o/s.
ae.
4) Var(W)
Ta) Use the craps principle. Imagine the following game: A tosses the biased coin once, then B.
tosses it once. If A's toss lands heads, then say that A wins the game; if A’s toss does not land
heads but B’s does, then say that B wins the game; otherwise the game ends in a draw. The
chance that A wins this game is p; the chance that B wins this game is gp: and the chance of a
draw is @?.
58»
4
8. ab)
@
Section 3.4
You can see that A and B are really repeating this game independently, over and over, until
cither A wins or B wins, and that the desired probability is P(A wins before B does). By the
craps principle, this is 5 = rf
Another way: In terms of T, the number of rolls required to produce the fist head, the desired
piobability is
PUD is odd) = PT =1) + PCT = 3) 4 PUP = 8) to pte ta Pt = hs
Apply th rans principe tothe situation in) the dasied probability ie P(B wine before A dos)
Fis = thy. OF compute P(T is even) = 38x. Or subtract the answer in a) from 1, because
Sheet the layers must see a head sometime
Apply the craps principle, this time with the gime consisting of A tossing the coin once, then B
tossing it ewie. The chance that A wins ths game jsp, while the chance that B wins is q= @?
Se the hans tas A ete he St Bead 7/9), andthe hans that D gt the Sat hed
~ @)/(1= 9"). Note that 1? = (I —9)(1 +a-+4?), 0 P(A gets the fest head) simplifies
Saeed gh Alteratvey in tees ofthe sendom arable T the chance the A gels the
fot head is the probability that T is of the form 1 4 3k for some # = 0, 1,2,.. similatly for B.
Solve slygr = 5 to get g = St and p = 254 = 381966.
ep is very amall, then the fact that A gets to toss frst doesn’t confer much of an advantage to
A. However, since B tosses twice as often as A does, you would expect that the chance that B
‘gets the frat head is clove to 2/3. Indeed, as q tends to 1:
P(A gets the frst head)
P(B gets the first head) —
Suppose that the player's point Xo is z = 4, 5, 6, 8, 9, or 10. Then P(win|Xo = z) is the
probability that in repeated throws of a pair of dice, the sum x appears before the sum 7
does. In the notation of the Example 2, imagine that A and B are repeating, independently, 2
‘competition consisting of the throw of a pait of dice. On each throw, if x appears, then A wins;
i€7 appears, then B wins; otherwise a draw results. The desired probability is the chance that
‘A wins before B docs, which, by the crape principle, is p23
= [2{[slaf{s{[o][7][s{o]{wlutn
Play) a/ae [ayse | a7se | a7o6 | 9/96 | 6736 | 9736 | «736 | 2736 | 2736 | 1796
FoiniXe=sy| o | 0 fas laolsul| 1 [slenol ae] 1 | 0
Plwin) = D2, PwinlXo
0 $43 36-24 ig 324 ir
9. Let X be the payoff. Then
PIX
‘That
E(X) = £(W?) = Var(W) +[E0W)}
Your
2) =)" fora = 1.2,
X = W*, where W has geometric (p) distribution on {1,2,...} for p = 1/2. Since W has
jp = 2 and vasience q/p? = 2.
net gain
sn by X 10, so B(net gain) = —4. That is, you are going to lose $4 per game
in the long run
10. a)
Condition on the first outcome and use the rule of average conditional probabilities: Let $=
(Gest trial results in success), F = (fest trial results in failure). Then
P(X = 2) = PUX = nl5)P(S) + P(X = al F)PLFY
Note that
PUX = nlS) = PWe
4g n> 2,
59Section 3.4
where Wr denotes the mumber of asses until the frst failure, which has a geometric distribution
on (1.2...) with parameter ¢. Similarly
P(X = n|F) = P(Ws =n —1)=4"*p,n 22,
Therefore
P(X =n) ato"s'p for n= 2,
2) Duplicate the method fr finding the moments ofthe geometric distribution: E(X) =
> ng™'p
(Lo.»)~1) +9 (LG.a) 1)
(csp -1) +9 -1)
Le d=1, where P(r) =D,
©) Similarly E(X*) = D2, n?p*!g + DS weap
Gen) +e(nea—y)
= 0 (ci8fs ~ 1) +0 (cigs 1)
= e+ ta.
Finally, use Var(X) = E(X?) -(£(X)P -
11, Write ga =1—pasg
a) P(A wins)
= P(A tosses H, B tosses T)+P(A tosses TH, B tosses TT)+P(A tosses TTH, B tosses TTT) +
= page + (qage)page + (gage) Page +
b) P(B wins) = ;2428-. (Interchange A and B.)
©) P(draw) = 1 = P(A wins) ~ P(B wins]
4) Let W be the number of trials (in which A and
(by either A or B). N has range (1, 2,3,
P(N = k) = Plfirst 1
both toss) required until at least one H is seen
. )- For k= 1,2,3,
ils see TT, kth tr
does not see TT)
= (gaga)*"(1 ~ gagn)-
Or: Bach trial is a Bernoulli tial with success corresponding to the event (at least one Hl
is seen among the tosses of A and B). The probability of success on each trial is then 1 —
P(no Hl is seen) = 1 qaqa. Therefore N, the waiting time until the Rrst success, is geometric
with parameter (I - gage).
12, Write gr = 1 = pisga
~m
a) PW, = We) = DZ, PWs = ks = 8)
= Siete = pBE = te
b) PIM < Wa) = DEL, PW