MTH5131: Actuarial Statistics – SOLUTIONS 2021
Hello, dear friend, you can consult us at any time if you have any questions, add WeChat: daixieit
MTH5131: Actuarial Statistics – SOLUTIONS
2021
Question 1 [12 marks]. [U]
(a) Previous national household longitudinal studies could be used for
pre-lockdown data. They could also be used for post-lockdown data, if they
(b) This is an inferential study.
(c) This study is longitudinal in nature. You would want to know how the alcohol,
e-cigarette and cigarette consumption of respondents to the study over a timespan that includes the lockdown.
(d) You might plot alcohol, e-cigarette and cigarette consumption over a timespan that includes the lockdown among different categories of respondents.
Question 2 [9 marks].
(a) We solve the equationS = = 9 →., 一2a2b4a一一c(b)20= 0
One solution is
╱ ←b(a) ╱ ←2(1)
We normalise to give
╱←
(2/3『
The other possible answer is
╱ 一(一)←
一2/3
(b) The eigenvalues all lie on a line in the Skree diagram ansd so they are all on the mountain.
The two largest eigenvalues only take 15/18 = 0.833 ì 0.90 of the total
variance. Either way, all components are principal.
Question 3 [16 marks].
(a) (i) The data is already ranked (although not in order from smallest to largest). The differences are We find that Eidi(2) = 22 and so
Candidate |
Interviewer 1 |
Interviewer 2 |
Difference |
A |
7 |
8 |
-1 |
B |
2 |
3 |
-1 |
C |
10 |
9 |
1 |
D |
14 |
14 |
0 |
E |
15 |
13 |
2 |
F |
6 |
5 |
1 |
G |
1 |
2 |
-1 |
H |
8 |
7 |
1 |
I |
9 |
10 |
-1 |
J |
4 |
4 |
0 |
K |
5 |
6 |
-1 |
L |
3 |
1 |
2 |
M |
13 |
15 |
-2 |
N |
11 |
12 |
-1 |
O |
12 |
11 |
1 |
rs = 1 一 一 1) = 0.9607
(ii) The t-statistic with 15 一 2 = 13 degrees of freedom is
(0.9607) ^13
^1 一 0.96072
The p-value 2P(Y ( 12.48077) = 1.303 × 10一8 is much less than 1% and is strong evidence to conclude that the rankings are correlated.
(b) We make a table of concordant and discordant pairs:
Rank1 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 |
Rank2 2 3 1 4 6 5 8 7 10 9 12 11 15 14 13 |
C 13 12 12 11 9 9 7 7 5 5 3 3 0 0 |
D 1 1 0 0 1 0 1 0 1 0 1 0 2 1 |
Totalling the columns gives nc = 96 and nd = 9 Thus
r = 14(一)/2 = 0.82857
Question 4 [18 marks].
(a) The density of Y is
fY (y) = 9y9 一1, 0 ì y ì 1.
The first moment is
E(Y) = (01 yfY (y) dy = (01 9y9 dy =
We set
y = E(Y) = → = 1 y一y
We find that
y = 0.92 +0.79 +0.90 +0.65 + = 0.8
Therefore
= 1 0.一0(8).8 一 1 = 4.
The likelihood is
910 │ yi ∶9 一1 .
The log-likelihood is
10
ì = 10 ln(9) + (9 一 1) E ln(yi) .
i=1
Taking derivatives:
+ ln(yi) = 0 → = 一
(Clearly, the second derivative is negative.) We get
= 4.116
(c) The likelihood is
910 │ yi ∶9 一1 (P(Y ì 0.60))2
We calculate that
P(Y ì 0.60) = (00.60 9y9 一1 = (0.60)9
The log-likelihood is
10
ì = 10 ln(9) + (9 一 1) E ln(yi) +29 ln(0.60)
i=1
Taking derivatives:
+ ln(yi) +2 ln(0.60) = 0 → = 一
(Clearly, the second derivative is negative.) We get
= 2.8976
extra information that some data were small relative to the other data.
Question 5 [12 marks].
(a) The expectation of th estimator is
E[C(Y1 +Y2 +Y3)| = C (E[Y1| +E[Y2| +E[Y3|)
= C ╱ + + 、
7C
=
4a .
In order for C(Y1 +Y2 +Y3) to be unbiased for 1/a we must have
C =
(b) The MSE of the the unbiased estimator we found in (a) is its variance:
C2 ╱ + + 、
= C2 = × =
Question 6 [9 marks].
(a) The prior is
f (p) o p4一1 (1 一 p)3一1 = p3 (1 一 p)2
The likelihood is
f (y|p) o I(p(1 一 p)yi) = p10 (1 一 p)E1 yi = p10 (1 一 p)42
i=1
The posterior is proportional to
p3 (1 一 p) 2 × p10 (1 一 p)42 = p13 (1 一 p)44
Therefore, the posterior is Beta(14, 45) distributed.
(b) The Bayesian estimate of p under square error loss is the expectation of the posterior. The posterior is Beta(14, 45) . Therefore, the Bayesian estimate of p under squared error loss is
14 14
14 +45 = 59
Question 7 [9 marks].
(a) E[s2 (9)| is estimated by the average of the sample variances:
E[s2 (9)| = = 81.25
The sample mean of the Xi’s is:
125 +85 +140 +175
and the sample variance of the Xis is:
1 3
4 一 1 E(Xi 一 X)2
(125 一 131.25)2 + (85 一 131.25)2 + (140 一 131.25)2 + (175 一 131.25)2
=
3
= 1389.583
Moreover,
1 3 1 1
Z = = 0.9883
(b) The credibility estimate of the amount per claim for the coming year for
product 3 is
(1 一 0.9883) × 131.25 +0.9883 × 140 = 139.8977
Question 8 [15 marks].
(a) In parameterised form, the linear predictors are (with i corresponding to the
levels of YO):
Model 1 : ai (2 parameters)
There are two parameters for the 2 combinations of YO.
Model 2 : ai+ βj (4 parameters)
There is one parameter to set the base level for the combination YO0, FMS0 and three additional parameters for the combinations of the higher levels of the two factors.
(b) The completed table, together with the differences in the scaled deviance and
degrees of freedom, is shown below.
Model |
Scaled Deviance |
Degrees of Freedom |
/ Scaled Deviance |
/ Degrees of Freedom |
1 YO YO +FMS |
40 30 23 |
5 4 2 |
10 7 |
1 2 |
(It is not necessary for the students to add the additional columns to the table.)
Comparing the constant model and Model 1
The difference in the scaled deviances is 10.
This is greater than 3.841, the upper 5% point of the x1(2) distribution. So Model 1 is a significant improvement over the constant model.
Alternatively, if we use the AIC to compare models, we find that since /(deviance)( 2 × / degrees of freedom, because 10 ( 2 × 1, Model 1 is a significant improvement over the constant model.
2022-07-29