INSTITUTE AND FACULTY OF ACTUARIES

EXAMINATION

30 September 2020 (am)

Subject CS2B – Risk Modelling and Survival Analysis

Core Principles
Time allowed: One hour and forty-five minutes

In addition to this paper you should have available the 2002 edition of the Formulae
and Tables and your own electronic calculator from the approved list.

If you encounter any issues during the examination please contact the Examination Team on
T. 0044 (0) 1865 268 873.

CS2B S2020 © Institute and Faculty of Actuaries

1 An actuary is investigating the asymptotic behaviour of the sample autocorrelation
function for two time series models.

(i) Generate, using a random number generator seed of 967, a simulated sequence
of n = 200 observations for a first-order moving average process with
parameter β1 = 0.4, assigning the simulated values to a vector called YMA. [2]

(ii) Generate, using a random number generator seed of 967, a simulated sequence
of n = 200 observations for a first-order autoregressive model with parameter
α1 = 0.45, assigning the simulated values to a vector called YAR. [2]

(iii) Plot, on four separate graphs, the sample autocorrelation function (sample
ACF) and sample partial autocorrelation function (sample PACF) for each of
the two time series models, YMA and YAR, generated in parts (i) and (ii). [6]

(iv) Comment on the general features of the plots in part (iii) with reference to
whether they are consistent with the theoretical behaviour of the
corresponding functions for the true models. [4]

The acf() function in R can also provide a vector output of the sample ACF values,
with component i giving the sample ACF at lag i – 1, provided that the plot argument
of the function is set to ‘FALSE’.

(v) Determine the numerical values for the sample ACF at lag 2, for each of the
two time series models, YMA and YAR, generated in parts (i) and (ii). [2]

(vi) Construct R code that

 first sets a random number generator seed of 967; and then

 generates 1,000 random vectors (of length n = 200) for each of the two
models in parts (i) and (ii); and
 assigns the values of the sample ACF at lag 2 for each random vector to
two vectors ACF2MA and ACF2AR (each of length 1,000).
[8]

(vii) Determine the mean and variance of the two vectors, ACF2MA and ACF2AR,
generated in part (vi). [3]

(viii) Plot, on two separate graphs, the histograms of the two vectors, ACF2MA and
ACF2AR, generated in part (vi). [4]

(ix) Comment on the results in parts (vii) and (viii), including whether they agree
with the expected asymptotic behaviour. [9]
[Total 40]

CS2B S2020–2
2 Before answering this question, generate the vector, X, in R using the following code:

set.seed(1027); X = rexp(n=1000, rate=0.01)

The vector X represents the gross claim sizes of 1,000 claims. The payments are to be
split between an insurance company and its reinsurer under an Excess of Loss
reinsurance arrangement with a retention level M = 400.

(i) Determine the proportion of the claims that are fully covered by the insurer.
[2]

(ii) Generate an additional vector, Y, which is of the same length as X, such that Y
represents the amounts to be paid by the insurer for each component of X. [1]

(iii) Generate an additional vector, 𝑍, which is of the same length as X, such that 𝑍
represents the amounts to be paid by the reinsurer for each component of X.
[1]

An actuary assumes that the underlying gross claims distribution follows an

exponential distribution of some unknown rate λ. The actuary needs to estimate λ
using only the claim amounts recorded in vector Y.

(iv) Construct R code that calculates the log-likelihood, as a function of the

parameter λ, given the claim amounts data in vector Y. [10]

(v) Determine the value of λ at which the log-likelihood function reaches its
maximum. [6]
[Total 20]

CS2B S2020–3
3 An analyst is investigating a life insurance portfolio data set that comprises two
variables, x1 and x2, for 200 policyholders. The analyst is exploring whether the 200
policyholders can be divided into two clusters (labelled A and B) based on the two
variables, x1 and x2.

Before answering this question, generate the data set, portfolio, in R using the
following code:

set.seed(2019);
portfolio = data.frame(x1=rnorm(200,3,1),
x2=scale(c(rnorm(70,4,1), rnorm(130,10,1))))

In the first stage of the investigation, the analyst decides to assign the first 100
policyholders in the data set to cluster A, and the remaining policyholders to cluster B.

(i) Construct a new column in the data set, portfolio, called group_label_stage1,
containing the policyholder cluster labels, defined above. [4]

(ii) Determine the coordinates (x1A , x2A ) of the centre of cluster A and the
coordinates (x1B , x2B ) of the centre of cluster B. [6]

(iii) Construct a new column in the data set, portfolio, called dist_A containing the
Euclidean distances between the policyholders and the centre of cluster A. [4]

(iv) Construct a new column in the data set, portfolio, called dist_B containing the
Euclidean distances between the policyholders and the centre of cluster B. [4]

The analyst decides to update the cluster labels by assigning to each policyholder the
label of the cluster whose centre is nearest, according to the distances calculated in
parts (iii) and (iv).

(v) Construct a new column in the data set, portfolio, called group_label_stage2,
containing the updated policyholder cluster labels, defined above. [4]

(vi) Generate a 2x2 matrix showing the number of policyholders with each
possible combination of values of group_label_stage1 and
group_label_stage2. [2]

(vii) Comment on the matrix generated in part (vi) with reference to how the cluster
labels have changed between group_label_stage1 and group_label_stage2. [4]

(viii) Plot the column, x1, of data set, portfolio, against column, x2, (with x1 on the
x-axis and x2 on the y-axis), using two distinct colours to identify clusters A
and B according to the label, group_label_stage2. [6]

The analyst decides to stop at this stage and to report group_label_stage2 as the final
set of cluster labels.

(ix) Comment on this decision. [6]

[Total 40]

END OF PAPER

CS2B S2020–4