Institute of Actuaries of India

Subject CT3-Probability and Mathematical Statistics

May 2008 Examination

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 examiner have given credit for any alternative approach or interpretation which they
consider to be reasonable.
IAI CT3 0508

1. a. Mean 140.4167; Median 128

b. The least two points are 113 and 116. The largest two points are 179 and 239.

So, the trimmed mean is [140.42*12-(113+116+179+239)]/8 = 129.75

c. Comment:
i. Mean gives equal weight to all the observations so it is more likely to get impacted by
the extreme observations.
ii. Median doesnt depend on the size of the observations but only on the order. So it
is least likely to get impacted by the extreme observations.
iii. Impact of outliers on the trimmed mean depends on the number of observations being
trimmed. In any case it is less impacted by extreme observations than the simple
mean.

d.
11S 2
X ~ F 112 , Now
1300
P ( S 2 ! 1300)
11S 2
P( ! 11)
1300
P ( X ! 11)
0.4433

Total [8]
2. P A1 A3 / A2  P A1 A2 A3  / P ( A2 )
P A1 / A2 A3 P A2 A3  / P ( A2 )
= P A1 / A2 A3 P A3 / A2 
= (0.4) (0.8) = 0.32
Total [2]

3.
18 1
P( A)
36 2

18 1
P( B)
36 2
4 1
P(C )
36 9

P A B C 
1
P ( A) P( B) P(C )
36

Page 2 of 10
IAI CT3 0508

However,
1
P( A B) z P( A) P( B)
6

1
P( B C ) z P( B) P(C )
12
1
P( A C ) z P( A) P(C )
36
A, B and C are not mutually independent.

Total [4]
1
4. G X (t ) E (t X ) 2 pq
x
| x|1
tx ;x r1,r2...

1 p
=
2 q x
t x q | x|
1 p
>  
= tq  t 1 q  t 2 q 2  t  2 q 2  ...
2 q
 @
=
p
2q
>
q t  t 1   q 2 t 2  t  2   ... @
p f
f r
q

r
= ( qt )
2q r 1 r 1 t
p
1
q q
qt (1  qt )  1 
1
=
2q t t
pq t 1 1
=  ; q | t |
2q (1  qt ) (t  q ) q

p t 1 1
=  ; q | t |
2 (1  qt ) (t  q ) q

Total [3]
1
e >log x 3.2 @
2
5. a. f ( x) /2
;x t 0
x 2S
= 0 for x < 0

b. EX = e P V e 3.21 / 2
2
/2
e 3.7


Var X = e 2 P V eV  1 = e7.4(e-1)
2 2

c. P(X > 8) = 1 P(X d 8). Since log X ~ N(P=3.2, V = 1),
log 8  3.2
P(X d 8) =
1
= (-1.12) , denotes the cdf of standard normal.

Page 3 of 10
IAI CT3 0508

= 0.1314
P(X > 8) = 1-0.1314 = 0.8686
Total [4]

6. a. Most suitable distribution for N is Binomial (90, p) where p is the probability of head.
Estimate of p is 50/90. Mean of N is 90*(50/90) = 50.
Variance is 90*(50/90) * (1-50/90) = 50*40/90 = 22.22.

b. N approximately follows Normal with mean 45 and variance 22.5. Using continuity
correction,
PBin (N > 50)
= PNor(N t 50.5)
(50.5  45)
= PNor ( Z t ) where Z ~ N (0, 1)
22.5
= PNor (Z t 1.16)
= 0.123

c. Using the above probability, the P-value of the test is 0.123. We do not have sufficient
evidence to reject Ho at 5% significance level.

d. Ho should be rejected for P-value < 0.05. Let n be the desired number of heads. Then
P(N t n) 0.05
(N 45) (n 45)
P( t ) 0.05
22.5 22.5
(n 45)
1.64
22.5
n 1.64* 22.5  45
n 52.77

So, Ho should be rejected for the number of heads t 53.

Total [7]

7. Xi~U(0,1)
1 1
EXi = , Var Xi = ; i = 1,2,,20
2 12
20
5
Given that Y = X i and hence Y ~ N 10,
i 1 3

Y  10 9.1  10
a. P>Y d 9.1@ P d
5/3 5/3
= P[Zd - 0.697]

Page 4 of 10
IAI CT3 0508

= )(-0.697) = 0.2423

8.5  10 Y  10 11.7  10
b. P>8.5 d Y d 11.7@ P d d
5/3 5/3 5/3

= P[ -1.162 d Z d 1.317]

= ) (1.317)  ) (1.162)
= 0.9061 0.1226 = 0.7835
Total [3]

8. a. P (Wait for ten minutes)

=P(0,1,2 passengers will arrive within the next ten minutes )

10 10 10
  10  10 1
= e 3  e 3 ( )  e 3 ( )2 0.3528
3 3 2!
b. Solving this part depends on the memoryless property of Poisson process. The waiting
period before the arrival of next passenger is exponentially distributed with an expected
value of three minutes.
P(having to wait another five minutes)= P (nobody arrives in the next five minutes)
5

3
=e 0.1883
Total [5]
T
kx
9. a. T
0
2
dx

T
k
T2 xdx
0

k x2 T
[ ]0
T2 2
k T2
*
T2 2
k
2
For the above integral to be 1, k should be 2.

b. The likelihood function is

n
L ( x, T ) 2 n T  2 n xi ; 0<xi  T
i 1

(Differentiating this wrt T and equating to 0 will not give a solution for T in terms of x).
L is a decreasing function of T
L is positive only when T t Max( xi )
So L is the maximum when T Max( X i ) ; which is the MLE of T.
Total [5]

Page 5 of 10
IAI CT3 0508

10. a. For Exponential O f ( x, O ) Oe  Ox and F(x) = 1- e  Ox ,x > 0

The likelihood function is
L(O , x) O7 * e  O ( 23 27 3952 688995) * e 300 O

b. The log-likelihood is
"( x, O ) 7 ln O  693O
w"( x, O ) 7
 693
wx O
Equating above to 0 gives MLE of O = 7/693 =.0101

c.
w 2 " ( x, O ) 7
 2
w x2
O
w " ( x, O )
2
7
So, E 
w x
2
O2
1 O2
CRLB for the MLE is = .000015 which is same as the
w 2 " ( x, O ) 7
 E
w x
2

asymptotic variance of the MLE

d. The 95% confidence interval for O is O r 1.96.S .D.(O ) .

Using the MLE of O and the variance the interval is:

(0.0101 1.96*0.0039) i.e.(0.0025,0.0177)

e. The likelihood is
L(O ; x, ) [1  e 25O ]3 * O7 * e  O [( 23 27 3952 688995)  7*25] * e [3003*25]O

Total [12]
11.a. The Scatter Plot
5

3
College grade

0
0.0 .5 1.0 1.5 2.0 2.5 3.0 3.5 4.0

High school grade

Page 6 of 10
IAI CT3 0508

b. D 0.346 ; E 0.825; r 0.778 ;

c. As the estimate of the slope is lower than 1, the expected performance in college is lower
than the performance in high school.

d. y D  Ex 0.346  0.825 * 2.1 2.0785 .

Actual college grade for a high school grade of 2.1 for observation 10 is much higher
than the expected. This observation can be treated as an outlier while fitting the
regression.
Total [11]
(n A * Y A.  n B * YB.  nC * YC . )
12. Y..
( n A  n B  nC )
8 *100  10 *110  9 * 95) 2755
Y.. 102.037
8  10  9 27

a. P Y.. 102.037

W A Y A.  Y.. 100  102.037 2.037

WB YB.  Y.. 110  102.037 7.963
WC YC .  Y.. 95  102.037 7.037
SSR 1075
V 2 44.79
n..  k 24

b. SS B n ( Y
i
i i.  Y.. ) 2 =8*4.15+10*63.41+9*49.52= 1112.98

SST = SSB + SSR = 1112.98 + 1075 = 2187.98

c. The ANOVA table is:

Source DF SS MS F
Between 2 1112.98 556.49 12.42
Residual 24 1075.00 44.79
Total 26 2187.98

F0.05, (2, 24) is 3.403. So there is enough evidence to reject the null hypothesis of equality of
all fertilizer effects.
d. The 95% confidence interval for P + WA is given by
V V
(YA.  t0.025, 24 , YA.  t0.025, 24 )
nA nA
44.79 44.79
(100  2.064 * ,100  2.064 * )
8 8
(88.44,111.56)
Total [10]

Page 7 of 10
IAI CT3 0508

13. a.6X = 378, 6Y = 333, 6X2 = 16112, 6Y2 = 1347, 6XY = 14340
1
6XY  X Y
r n
1
n
6X 2  X   2 1
n
6Y 2  Y  2

= 0.679514 = 0.680 (App)

H0 : U = 0 ; H1 : Uz 0

r n2
Test statistic is t
1 r2
0.68 7
= = 2.453
1  0.68 2

The table value of t distribution for 7 df at 5% level is 1.895.

Conclusion: Reject H0.

b. H0 : V 12 V 22 ; H1 : V 12 z V 22
(Usually the larger variance is taken in the numerator)
X 42 ; s X2 29.5 and Y 37 ; sY2 143.75

Fcal = S Y2 / S X2 4.87 and Ftab (8,8) = 3.44

Conclusion: Reject H0

V Y2 1 sY2 sY2
c. 90% confidence interval for is , F
F (8,8) s 2 D / 2 (8,8)
V X2 D /2 X s X2
= [1.416,16.757] , F0.5 (8,8) 3.44
(The CI for V / V 2
X
2
Y [.0596, 0.7059]

d. Changes in Scores (d): 5 -7 -11 8 - 11 21 0 -9 1

(d = Y-X)
d 5 ; s d 9.206 and n= 9.
H 0 : Pd = 0 Vs H1 : Pd >0

d 5
The Test statistic is : t 1.629
sd 9.206

n 3
Table value t0.05(8) = 1.86; Conclusion: Do not reject H0

Page 8 of 10
IAI CT3 0508

e. 95% confidence interval for the difference between the two population means (Py- Px) is
s
{ ( P y  P x ) r (tD / 2 , n 1 ) d }
n
which simplifies to (-12.077,2.077).
( The confidence interval for (Px- Py) will therefore be (-2.077, 12.077).)
Total [14]

14 a. Note that value of k in the density does not affect the conditional expectation.
E(X/Y=y):
f
f Y ( y) f ( x, y)dx
f
Y
= k ( x  10)( y  10)dx
10

=k (y-10)3 ; 10 < y < 30

2( x  10)
Hence, f X / Y ( x / y ) ; 10 < x < y
( y  10) 2
= 0; otherwise
2( x  10)
y

E( X / Y y) x dx
10 ( y  10)
2

2
= 10  ( y  10) ; 0 < y < 30
3

2( x  10)
20
b. P ( X ! 20 / Y 25) (25  10)
0
2
dx =5/9

Total [7]

15. The pmf of X

n x
n 1 1
x

p X ( x)
x 2 2
n 1
n

= ; x 0,1,2,...n
x 2
n
and P ( N n) ; n 1,2
3
P  X x, N n  P  X x / N n P( N n)
n 1 n
n

= ; x 0,1,2,...n; n 1,2
x 2 3

Page 9 of 10
IAI CT3 0508

Hence, P ( X x) P X
n
x, N n

Then, P X 0 P X 1  P X 2 =
1 1
0, N 0, N  1/ 3
6 6
P X 1  P  X 2
1 1
P( X 1) 1, N 1, N =  1/ 2
6 3
P ( X 2) P  X 2, N 2  1 / 6
Mean of X: 5/6
Variance of X: 17/36

Total [5]

***************

Page 10 of 10