01.06.2015 Views

Actuarial Modelling of Claim Counts Risk Classification, Credibility ...

Actuarial Modelling of Claim Counts Risk Classification, Credibility ...

Actuarial Modelling of Claim Counts Risk Classification, Credibility ...

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

<strong>Risk</strong> <strong>Classification</strong> 83<br />

T 2 =<br />

∑ n<br />

i=1<br />

√<br />

∑n<br />

i=1<br />

(k i −̂ i 2 − k i<br />

)<br />

(<br />

k i −̂ i 2 − k i<br />

) 2<br />

<br />

and<br />

T 3 = √<br />

1<br />

n<br />

∑ n<br />

i=1<br />

(k i −̂ i 2 − k i<br />

)<br />

∑<br />

(<br />

) 2<br />

√ <br />

n ∑n<br />

i=1̂ −2<br />

i k i −̂ i 2 − k i i=1̂ 2 i<br />

All these test statistics are or0 1 distributed. For Portfolio A, T 1 = 918, T 2 = 613 and<br />

T 3 = 438. All the p-values are less than 10 −4 leading to the rejection <strong>of</strong> the null hypothesis<br />

(and thus the Poisson model) in favour <strong>of</strong> the mixed Poisson model.<br />

2.5 Negative Binomial Regression Model<br />

2.5.1 Likelihood Equations<br />

Overdispersion is taken into account by the inclusion <strong>of</strong> a random effect, representing<br />

an unknown relative risk level. More precisely, assume that 1 n are independent<br />

ama a distributed random variables, i.e. the common probability density function<br />

<strong>of</strong> the i s is given by (1.35). In this case, E i = 1 and V i = 1/a.<br />

Note that the assumptions made about the i s are rather strong. Their common<br />

distribution means that the effect <strong>of</strong> hidden variables does not depend on observable<br />

ones.<br />

Conditional on the observable characteristics summarized in the vector x i and on the<br />

random effect i = , the annual claim number caused by policyholder i conforms to the<br />

oi i law. In other words, i is the expected claim frequency for policyholder i (based<br />

on x i ) and is the relative risk level for this policyholder (if

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!