Академический Документы
Профессиональный Документы
Культура Документы
INDICATIVE SOLUTION
Introduction
The indicative solution has been written by the Examiners with the aim of helping candidates. The
solutions given are only indicative. It is realized that there could be other points as valid answers and
examiners have given credit for any alternative approach or interpretation which they consider to be
reasonable.
IAI ________ CT3 0511
Let x = z a.
since
1 for all x [ 0, ) and a > 0
[Total: 5]
Q2 Let X denotes the variable stating the gross income in units of 1,000.
321.6
We are given:
10628.31
* 32.16
This implies:
+ " 5.6338
*
Let Y denote the variable stating the income net of income tax in units of 1,000.
Note this formula holds only if no gross income is lower than 10,000. We assume this
holds for rest of the computations.
This means:
Page 2 of 18
-. 0.7 0 * 1 3
IAI ________ CT3 0511
0.732.16 1 3
25.512
+2 0.7 0 +
0.75.6338
3.944
Thus, the sample mean of the 10 net incomes is 25,500 while the standard deviation of the
same is 3,900. (The numbers are rounded to nearest hundred as specified in the problem)
[Total: 3]
Q3 As R has values in (0, 1), the density function 56 7 is non-zero only for r (0, 1).
1 7 1 1
?; ,; @ 1 ?; , ; A @
2 711 2 711
1 E1 F G
B
B B
[C X ~ U(0, 1)]
27
711
56 7 86 7
H
HB
2 27
F
7 1 1 7 1 1
2
7 1 1
[Total: 5]
Page 3 of 18
IAI ________ CT3 0511
Given Y =2 we have
2 T QU
S
S
Thus,
K
P I : > : > 5M7 I 0, 1, 2
U U
Now,
; P
KV ; P I
KV ; I P I + ; D P 7
WD
Q
D
DX
K
:U> :U>
J
KV K!
YZU[
J
KV
U K!
+
\WX]
Z J Z
U
U KV K! U
U
U
[Total: 5]
Page 4 of 18
IAI ________ CT3 0511
0.75 1 0.00005.
0 5M7 ` 0
0.8 5M7 0 ` 1000 a
Thus, 8X _
0.75 1 0.00005 5M7 1000 5000
1 5M7 5000
By inverse transformation method we will have for u generated randomly from U (0, 1):
0 W5 0 b ` 0.8a
8 b c
20000 b F 0.75 W5 0.8 b 1
The random samples for the first five trials are 0.95, 0.65, 0.75, 0.55 and 0.85.
b 0.65 T 0
bU 0.75 T U 0
bZ 0.55 T Z 0
[Total: 3]
Page 5 of 18
IAI ________ CT3 0511
Q6
(c) The distribution of each Xi is symmetric and so the distribution of the sum d feV ;e is
also symmetric.
[Total: 4]
Q7 Note here the random variable X takes values at discrete points and continuous regions
with probability greater than zero.
Let P[ X < 1 ] = c
For x < 1,
8 |; ` 1
h h
i=j k
For 1 < x 2,
F 1 8 |; 1
hh
i=n
Then,
d 1 F 8 o 1 F m
5M7 0 ` 1a
0.75 F m2 F 5M7 1 2
Page 6 of 18
IAI ________ CT3 0511
p; d
q 1 F m 1 q 0.75 F m2 F
UUk
Z
[Total: 5]
199.5 F 1.96
Solving for 1:
{ F 402.8416 1 39800.25 0
402.8416 55.50094
{ 173.67 M7 229.17
2
200.5 F F1.96
Solving for 2:
Page 7 of 18
{
F 404.8416 1 40200.25 0
IAI ________ CT3 0511
404.8416 55.6392
{ 174.60 M7 230.24
2
Thus, the approximate 95% symmetrical confidence interval for will be (173.67, 230.24).
[Total: 7]
We have 5 ; 0 ` ` , 0
Q9
ln 5 ln F 2 ln 1
F
f
F 1
f
Thus, p E G F 1
f
F 1
U
F 1 F
U U
.
, 9
Here, 9
f
U
f
.
U
f
Thus the asymptotic variance of the maximum likelihood estimator of is .
[Total: 5]
Page 8 of 18
IAI ________ CT3 0511
Q10 We have the random variable X with a density function:
5 | 1 F
U
0 implies:
H
(a) The mode of this distribution will be that value of x where the first derivative of f(x)
H
vanishes. Computing this derivative the expression
5
(Ignoring constant factors)
F 1 1 F 1 2 F 1 1 F F1 0
1 F 2 F 3 1 F 1 0
This can be solved for xm that makes this an equality and gives
++bWD|
U U
One can check that at this value of x, the second derivative is indeed negative.
(b) The most powerful test will be given by the Neyman-Pearson Lemma using the Likelihood
Ratio technique. It states that if C is a critical region of size and there exists a constant k
` m7XWm b
eKeH fHB
eKeH fHB
21 F 1
201 F U 101 F
Page 9 of 18
IAI ________ CT3 0511
(c) X has been observed as 1/7.
Here given the sample observed we are trying to compute the exact shape of such a critical
region so that it can be used to compute the p-value as desired in part (d).
Looking at the critical region we obtained in (b), the test should take the form:
1 1
Reject H0 if
;1 F ; F ?1 F @ 0
7 7
36
} ;1 F ; F 0
343
1 4 9
} ?; F @ ?; F @ ?; F @ 0
7 7 7
9
mX W5 ; : , >
Z
Thus we get:
1 4
9
mX ? ` ; ` @
7 7
Z
q 21 F
27
49
[Total: 10]
Page 10 of 18
IAI ________ CT3 0511
Q11 The moment generating function of X
X p
=
q
5
1
q
1 q
2
E 1 G
for |t| > 1
Now,
Thus,
p; 0 0
" 0 F 0
2
[Total: 5]
Q12
(a) The likelihood function is
Page 11 of 18
(b) Let Y = max ef ;e
IAI ________ CT3 0511
8R - P -
For 0 y
E max ;e -G
ef
; - ; - ;f -
f
;e -
eV
f 2
q 2
eV
f
2
2 f
Thus,
52 - : > 5M7 0 -
H f2
H2
2D - f
2D f 2D
pY [ q - : > - q - -
-
f 2D 1 1
pY [ F F
f
2D - f
2D D
pY [ q - : > - q f - f -
- D11
[Total: 8]
Page 12 of 18
For simplicity, it is assumed that the decisions ; , ; , ;f of the different voters are iid
IAI ________ CT3 0511
Q.
Q13
f f
f 7;f Q 1 F Q
We have
EOf F , Of 1 G
Thus,
f f
contains the 95% confidence interval for .
3%
If we want the error to be 3%,
f
then
n: > 1112
.U
[Total: 3]
(a) The Neyman-Pearson theory states that the most powerful test for against would
Q14
be that one which for given = 0.3 and critical region C which maximises ;
Therefore, the most powerful test is the one which rejects if X=1 or 3
The power of this test = ; 1 M7 ; 3
0.3 1 0.3 0.6
[Total: 5]
Page 13 of 18
IAI ________ CT3 0511
Q15
5 O, ln; 1
1FO ln; F O
\WX] ~0, 1
11
For j = 2, 3 9:
5 O, ` ln;
2 2
11
ln; F ln;
2 2
11
FO FO
2 F 2
~
F
5 O, ln; 5
5FO
1 F
^
Given that = 3, we observe that the intervals are symmetric around 3. So, we can say:
E6 = E5 = 267.057
E7 = E4 = 140.229
E8 = E3 = 37.125
E9 = E2 = 5.202
E10 = E1 = 0.387
Observe
V p 900 which is as expected.
Page 14 of 18
IAI ________ CT3 0511
(c) The chi-square goodness-of-fit test is to test:
H0: Data is a sample from the model M
v/s H1: Data is not a sample from the model M
Observed = Oi
Expected = Ei
Chi-Square =
The expected numbers in the first and the last group are too small; hence we combine them
individually with the immediate next group so that the expected frequencies for all cells is
at least 5.
[Note: Combined two pairs of groups & no estimation of any estimator is required]
10.618 F 10.5
By linear interpolation:
8 10.618 0.8380 1 ? @ 0 0.8614 F 0.8380 0.84352
11 F 10.5
(d) Here, we would have estimated the two parameters and . This means we will lose 2 more
degrees of freedom and d.f. = (10 2) 1 2 = 5.
Page 15 of 18
IAI ________ CT3 0511
Here the p-value has significantly decreased compared to earlier circumstance and at this
value there might be some evidence that the null hypothesis might not hold.
[Total: 10]
Q16
(a) The variable x takes values 1, 2 . 30. Sample size n = 30.
d e F *
e F D*
DD 1 12D 1 1 DD 1 1
FD
6 2D
DD F 1
12
30 30 F 1
12
2247.5
[Alternatively, this can be derived from first principles]
(b)
y F y
Yd22 F d2 d [
f
Page 16 of 18
IAI ________ CT3 0511
(d) We have SS S
Hence SS S F SS 1020.6 F 860.48 160.12
X Y F [+
Y [
We know the test statistic
We have n 2 = 28 degrees of freedom. From tables: X.^ 2.048 and X.^ 2.763.
We may not have sufficient evidence to reject Ho under 5% significance but there may be
some evidence to do so under 1% significance. Note, we got R2 = 15.7% which indicates only
15.7% of the variation is explained by the model meaning the linear relation is sufficiently
weak.
[Total: 10]
dd6
ii. Estimate for common underlying variance in serotonin level =
fK
Page 17 of 18
IAI ________ CT3 0511
Here: SSR = SST - SSB
1
dd -e
F -.. 676 1 382 1 78 1 196 F 17027
D
261.630
1 1 66 46 20 38 170
dd -e. F -.. ~ 1 1 1 F
De D 7 6 6 8 27
1222.12 F 1070.37
151.749
dd6 261.63 F 151.749 109.88
Hence 4.777
.
Z
(b) ANOVA
5% critical value for 8U.U is 3.028 << 10.589, hence we have sufficient evidence to
reject at the 5% level. We conclude that there are differences between the mean
levels of serotonin.
[Total: 7]
*******************
Page 18 of 18