Institute of Actuaries of India

Subject CT3 Probability & Mathematical

Statistics

May 2010 Examinations

INDICATIVE SOLUTIONS

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 Indicative Solution CT3 0510
 (a) Arranging the data in ascending order

199, 201,203,206,209,210,210,212,215,215,216,218,218,230,240
Here,n = 15.
One can use,
= = 206.75

= = 217.50
Inter Quartile Range = .

Other can use,

= = 206.00
= = 218.00
Inter Quartile Range = .

Hence, both statisticians A and B are correct but their ranges differ due to use of two different
definitions of quartiles. 

 (b) sample mean =

sample variance = 

 (c) X N(213.47, 114.70)
230 2
213.47
P(X> 230) = P(Z> ) = P(Z > 1.543)
114.7
Using the tables and using linear interpolation between 1.54 and 1.55, we get the values as
P(Z > 1.543) = 1 0.93846 = 0.06154

 (d) Using empirical distribution, we have 15 observations and only one observation is greater
than 230. Hence probability = = 0.067.
Thus, we can see that both the probabilities are close.  [9]


 We use the formula,
Let X be the random variable representing the mass of an elephant.

We know that
Pr(X = 900) =
Pr(X = 1,000) = , and
Pr(X = 1,396) =

Page 2 of 12
IAI Indicative Solution CT3 0510
E(X) = 900 + 1,000 + 1,396 = 1,057.33

Var(X) = 2 = (900- 1,057.33)2  + (1,000-1,057.33)2 + (1,396-1,057.33)2  40,083.56

3 = Var(X)1..5 = 80,25,081.09

3 = (900- 1,057.33)3  + (1,000-1,057.33)3 + (1,396-1,057.33)3  = 80,25,685.58

Thus, = 1.00 [3]


 D = Has H1N1
T = Test result is positive
Given: P(D) = 0.001 so P(Not D) =0 .999
P(T | Not D) = 0.05
P(T | D) = 1.00
We need to find P(D | T)
Now, P(TD) = P(D)P(T | D)
= 0.001
P(T) = P(TD) + P(TNot D)
= P(TD) + P(Not D)P(T | Not D)
= 0.001 + (0.999)(0.05)
=0 .05095
P(D | T) = P(TD) /P(T) = 0.001/0.05095 = 0.019627 [4]

 Y = e2X + 5 implies that
FY(y)
= Pr(Y y)
= Pr(e2X + 5 y)
= Pr(e2X y-5)
= Pr(2X )
= Pr(X )
= ]
= . [2]

 C X (t ) lln M X (t )
'
'
M X (t )
C X (t )
M X (t )

Page 3 of 12
IAI Indicative Solution CT3 0510

Now M X (t ) E (e tX )
Expanding the series, we have
t2 2
=> M X (t ) E (1 tX X .....)
2!
t2
=> M X (t ) 1 tE( X ) E ( X 2 ) ...
..
2!
' 2
=> M X (t ) E ( X ) tE( X ) ..
...
'' 2
=> M X (t ) E ( X ) ..
...

Now,
'
M X (0) 1 M X (0) E( X )
;
'' 2
M X (0) E( X )
Using the above values, we get

2 2
''
E ( X ) *1 ( E ( X ))
=> C (0) V
Var[ X ]
X 1
[4]


 We can define the probability density function (pdf) of B as a system of two equations with two
unknowns:

At x = 8, we have,
We also know that the area underneath a pdf must equal 1.
x
-
Thus, 8c + ke 7 /7dx = 1
8

Page 4 of 12
IAI Indicative Solution CT3 0510
=> 8c + -ke-x/78 = 1
=> 8c + ke-8/7 = 1. But c = ke-8/7/7, so ke-8/7 = 7c, and therefore 8c + 7c = 1 and so c = 1/15.
Over the interval 0 x 8, the area underneath the pdf is 8c = 8/15 [5]

 We first find = 0.23(2) + 0.34(5) + 0.43(7) => = 5.17.

Now we find 2 = 0.23(2-5.17)2 + 0.34(5-5.17)2 + 0.43(7-5.17)2 = 2 = 3.7611.

2
(Xi X)/n
Now Sn2 = = (n-1)Sn2/n
E(Sn2 ) = E((n-1)Sn2/n) = (n-1)E(Sn2)/n = (n-1)2/n

So for n = 13, E(S132) = 122/13 = 3.4718.

Thus, the bias of Sn2 is E(Sn2 )- 2 = 3.4718 - 3.7611 = -0.2893 [3]


 x
(x )

(a) E ( X ) e dx

let y = x - ; thus
y
y
E( X ) e dy
0

y y
y 1
e dy + e dy
0 0

=+
2 (x )
2 x
Now, E ( X ) e dx

let y = x - ; thus
2 y
2 (y )
E( X ) e dy
0

2 y y y
y 2 1 y
e dy + e dy + 2 e dy
0 0 0
2 2 2 2
= 2 + + 2 = + ( + ) .
=

(b) We use two moments, since there are two parameters to estimate:
We are given that there are 50 pieces of data and that their sum is 8200. Thus,

Page 5 of 12
IAI Indicative Solution CT3 0510
E(X) = 8200/50 = 164. Likewise, E(X2) = (xi2)/50 = 40000.

Thus, we set up the following system of equations:

A: 164 = + ;
B: 40000 = 2 + ( + )2.

We know that ( + )2 = 1642 = 26896, so 2 = 40000 - 26896 = 13104, and thus = 114.47
Hence, = 164 - = 49.52. [7]


 Let X be the outcome from the one dice rolled. Since N is determined by the value of X,
E(N) = E(X) and Var(N) = Var(X).

First, we find E(X) = (1/8)(1+2+3+4+5+6+7+8) = 4.5 = E(N)

Now we find Var(X) = (1/8)((1-4.5)2 + (2-4.5)2 + (3-4.5)2 + (4-4.5)2 + (5-4.5)2 + (6-4.5)2 + (7-4.5)2 +
(8-4.5)2) = 5.25 = Var(N)

Let S be the payoff in each individual round. Then

Var(S) = E(N)*Var(X) + Var(N)*E(X)2 = 5.25(4.5 + 4.52) = 129.93. [4]


 First, we find the mean loss per insured. 

Let N = frequency of losses to individual insured; X = magnitude of losses; and S = total loss to
individual insured.
E(S) = E(N) * E(X) = 0.1 * 500 = 50.
For n insureds, the expected loss is therefore 50n, and 1.5 times the expected loss is 75n.
Now we find the variance of the loss per insured.
Var(S) = E(N) *Var(X) + Var(N)* [E(X)]2
= 0.1*(1000)2 + 0.1*(500)2
=0.1*1,250,000
The variance for n insureds is thus 125000n, and the standard deviation is 353.55(n).

We want to find n such that Pr(Total loss 75n) <=0.05.

For minimum n , we need Pr(Total loss 75n) =0.05

By the central limit theorem assuming total losses follow normal distribution
Pr(Total loss 75n) =0.05
Pr[z (75n - 50n)/(353.55(n))] =0.05

Page 6 of 12
IAI Indicative Solution CT3 0510

1 - ((75n - 50n)/(353.55(n))) = 0.05

(0.07071(n)) = 0.95
0.07071(n) = 1.645
n = (1.645 / 0.07071)2 = 541.2,
which we round up to 542 insureds. [5]


 MSE = E[(X2 2)2]
= E(X4 2X22 + 4)
= 244 2(22)2 + 4 = 214 (using E(Xr) = (1+r) * r from the tables)
Any alternative solution is equally acceptable
[3]

 Next year, the severities will change to 0.6 of their previous amounts; thus, the severities will
become
0.6*350 = 210, 0.6*460 = 276, 0.6*900 = 540, 0.6*3400 = 2040.
With the 200 deductible, the payout for each magnitude of loss will be
10, 76, 340, and 1840, respectively.

Thus, the expected payout per claim is 10*0.4 + 76*0.3 + 340*0.2 + 1840*0.1 = 278.8.

Total expected payout is E(N)*(Expected payout per claim) = 5667*278.8 = 1,579,959.6.

[3]

 (a) If we have m values between y and z and p values above z in addition to the n known values,
then the likelihood function is given by

Thus, using above, if we have 9 losses less than 10, 6 losses between 10 and 25 and 5 losses above
25, then the likelihood function is given by:
9 6 5
L( ) [ P( X 10
1 )] * P(10 X 25
2 )] * [ P( X 25
2 )]
L() = (1 /10)9 * {(1-/25) (1-/10)}6 * {(1-(1-/25)}5
Which is proportional to (10 )9 * 11

Log{L()} = 9Log(10 ) + 11 Log()

Differentiating w.r.t

= -9/(10 ) + 11/

Page 7 of 12
IAI Indicative Solution CT3 0510
Equating this to zero, we get
110 = 20

= 5.5

Differentiating w.r.t again, we get

= -9/(10 )2 - 11/ 2

We can see that is always negative, which implies = 5.5 is the MLE of

(b) Substituting = 5.5, we have = -0.80808

CRLB = or

CRLB = 1.2375, as the = 5.5

(c) N(, CRlb)

95% CI is 1.96 CRlb = (3.32, 7.68) [10]


 For the Poisson distribution, the mean, , is estimated as 230/1000 = 0.23. 
2
Number of Poisson Expected Observed (O-E) / E
days work Probability Number of Number of
missed Workers Workers
0 0.794534 794.53 818 0.69
1 0.182743 182.74 153 4.84
2 0.021015 21.02 25 0.76
3 or more 0.001708 1.71 4 3.08
2
Total 1 1000 1000 = 9.37
Marks 1.5 marks 2 marks

The 2 distribution has 2 degrees of freedom because there are four categories and the Poisson
parameter is estimated (d.f. = 4 1 1 = 2)

The critical value for a chi-square test with two degrees of freedom and 5% level of significance is
5.99. Since Chi Square calculated is greater than Chi Square tabulated, null hypothesis is rejected.
[5]


Page 8 of 12
IAI Indicative Solution CT3 0510
 (i) Missing entries are:
d.f. between regions = 3-1 = 2
d.f. total = 30-1 = 29
d.f. Residual = 27 by subtraction
SS = 5992.8 by subtraction
MSS = 2176.4 being 4352.8/2.
MSS = 221.96 being 5992.8/27.
Observed F = 9.8056 on 2 and 27 d.f.
From tables the 5% critical point is 3.354
Significant at 5%, so there is quite strong evidence of a difference in the mean claim amounts for
the three regions (significant at even 2.5% or 1% critical point, so answer do not change)
(ii) 95% confidence interval for ( A B
) is

1 1
( y A y B ) t 5%
2
, 27 10 10
1
giving
1 1
(155.48 164
1 .68) 2.052 221.96 as = MSS
10 10
1
-9.2 13.67 = (-22.87, 4.47)
The interval contains zero indicating no difference between the underlying means for regions A
and B.
The significant result of the F-test clearly comes from region C mean being much lower than the
region A and B means [7]
 (a)

60

50

40

y 30

20

10

0
0 10 20 30 40 50 60 70
x

Page 9 of 12
IAI Indicative Solution CT3 0510
(b) (i)
Total
x 0 10 20 30 40 50 60 210
y 16 27 28 39 39 48 51 248
xy 0 270 560 1,170 1,560 2,400 3,060 9,020
x2 0 100 400 900 1,600 2,500 3,600 9,100

Sxy = xiyi xi yi /n- = 9020 210 * 248 / 7 = 1,580

Sxx = xi2 (xi)2/n = 9100 - 2102 / 7 = 2,800

b = Sxy / Sxx = 1580 / 2800 = 0.564

a = y-- - b x-- = (248 / 7) - 0.564 * (210 / 7) = 18.5

y = 18.5 + 0.564x

b(ii) Draw a line joining the two (x, y) co-ordinates: (0, 18.5) and (60, 52.4)
{OR any other co-ordinate provided the value of y is correctly calculated and plotted, against
the value of x}

(c) (i) 2 = (Syy S2xy/Sxx) / (n 2)

Syy = yi2 ny2 = 9716 2482 / 7 = 929.71
Thus, 2 = (929.71 15802 / 2800) / 5 = 7.63

se( b ) = (2 / Sxx)1/2 = (7.63 / 2800)1/2 = 0.0522

95% confidence interval for b is b {t0.025, 5 * se( b )}

= 0.564 (2.571 * 0.0522)
So 95% confidence interval is (0.430, 0.698)

c(ii) The 95% two-sided confidence interval does not contain the value 1, so the two-sided test
conducted at 5% level results in H0 being rejected.

(d) (i) Difference between 8:40 am and 7 am is 100 minutes.

z = Time available Time taken

= (100 x) y

= (100 x) ( a + b x)
= 81.5 1.564x

Page 10 of 12
IAI Indicative Solution CT3 0510

(ii) z = 81.5 1.564x

For z = 0, x = 81.5 / 1.564 = 52.11, taken equal to 52 minutes as approximation.

The latest time is therefore 7:52 am.
[14]

 H0: there is no difference in the I.Q before and after the training programme
H1: there is difference in the I.Q before and after the training programme
I.Q before I.Q after
Candidate d = (y-x) d- d (d - d )
2
training (x) training (y)
1 110 115 5 2.5 6.25
2 105 117 12 9.5 90.25
3 111 109 -2 -4.5 20.25
4 125 124 -1 -3.5 12.25
5 120 122 2 -0.5 0.25
6 115 114 -1 -3.5 12.25
Total d =15 0 2
(d - d ) = 141.5

d = 15/6 = 2.5

(d d )2 141.5
s = = 5.32
n 1 5
d
under H0 , ~ tn 1
s
n

thus,

Since < , H0 is accepted. Hence, there is no difference in the I.Q before and
after the training programme

1 1 r 1 1
(b) We know that loge ( ) is approximately normal distributed with mean loge ( )
2 1 r 2 1
1
and standard deviation
n 3
where r is the sample correlation coefficient and is the population correlation coefficient
1 1 r 1 1 1
95% C.I. for loge ( ) is given by loge ( ) 1.96
2 1 r 2 1 n 3
1 1 1 1 r 1
=> loge ( ) lies between loge ( ) 1.96
2 1 2 1 r n 3
1 1 1 1.7
=> loge ( ) lies between loge ( ) 1.96*(1/7) = 0.5873, 1.1473
2 1 2 0.3
=> lies between 0.528 and 0.817

Page 11 of 12
IAI Indicative Solution CT3 0510
1
(c) f(x, y) = (2 X Y ) X = 0,1,2
27
Y = 0,1,2
Joint probability distribution f(x, y) of X and Y is given by table
X/Y 0 1 2
0 0 1/27 2/27
1 2/27 3/27 4/27
2 4/27 5/27 6/27
fy(Y) 6/27 9/27 12/27

The marginal probability distribution of Y is given by f y ( y ) f ( x, y )

x

The conditional distribution of X given Y=y is given by the table below

f xy ( x, y )
f X /Y (X x /Y y)
f y ( y)
X=x/Y=y 0 1 2
0 0 1/9 2/12
1 2/6 3/9 4/12
2 4/6 5/9 6/12
 [12]

Total Marks 100

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

Page 12 of 12