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

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

<strong>Risk</strong> <strong>Classification</strong> 75<br />

used). The ratio <strong>of</strong> the deviance to the number <strong>of</strong> degrees <strong>of</strong> freedom should be close to<br />

1 to indicate goodness-<strong>of</strong>-fit. Here, we obtain 05357 (the deviance is equal to 7764.83).<br />

Note however that the data should be grouped to make the Chi-square approximation more<br />

reliable, so that we cannot use this statistic at the present stage.<br />

An important aspect <strong>of</strong> insurance ratemaking with generalized regression models is the<br />

selection <strong>of</strong> explanatory variables in the model. Changes in goodness-<strong>of</strong>-fit statistics are<br />

<strong>of</strong>ten used to evaluate the contribution <strong>of</strong> subsets <strong>of</strong> explanatory variables to a particular<br />

model. One strategy for variable selection is to fit a sequence <strong>of</strong> models, beginning with<br />

a simple model with only an intercept term, and then include one additional explanatory<br />

variable in each successive model. The importance <strong>of</strong> the additional explanatory variable<br />

can be measured by the difference in fitted log-likelihoods between successive models.<br />

Asymptotic tests computed by the GENMOD procedure enable the actuary to assess the<br />

statistical significance <strong>of</strong> the additional term (this is called Type I analysis in SAS R ).<br />

Another strategy (adopted here) consists <strong>of</strong> starting from a model incorporating all the<br />

available information, and then excluding the irrelevant explanatory variables. To this end,<br />

the GENMOD procedure generates a Type 3 analysis (analogous to Type III sums <strong>of</strong> squares<br />

in the GLM procedure). A Type 3 analysis does not depend on the order in which the terms<br />

for the model are specified (in contrast to the Type 1 analysis).<br />

Type 3 analysis compares the complete model (that is the model which includes all the<br />

specified variables) with the different submodels obtained by deleting one <strong>of</strong> the explanatory<br />

variables. It enables the actuary to test the relevance <strong>of</strong> one variable taking all the others<br />

into account. It roughly corresponds to the backward approach: at each step, we exclude the<br />

variable with the largest p-value until no more can be excluded (i.e. until all the p-values<br />

are smaller than a fixed threshold, generally 5 %). Note that the Type 3 analysis works with<br />

the variables and not with the levels <strong>of</strong> these variables. Indeed, it is possible to obtain a<br />

relevant variable for which some levels are not relevant. The results <strong>of</strong> the Type 3 analysis<br />

are as follows:<br />

Source DF Chi-square Pr > Chi-sq<br />

Gender ∗ Age 7 7516

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

Saved successfully!

Ooh no, something went wrong!