Page 23 of the GLM chapter says the response variable of the claim frequency model is the number of claims.
So I would presume that if for a given row i I had 1 claim in a six month period than Yi = 1.
but, when I look at the linear predictor I have X.B where neither the Xs or the Bs appear to adjust for the fact I am only looking at a six month period (B is a one dimensional matrix, independent of i) , thus my µi is number of claims in one year, claim frequency; not number of claims.
so if:
Yi = E[Yi] + ε = ui + ε
I have my Yi as a six monthly rate, and ui as an annual rate, which does not make much sense to me.
...or is it the case that X is not a matrix of 1s and 0s, as shown in the examples, and row Xi would read: 1/2 male, 0 female, 1/2 age 35, 0 age 36, etc?
Last edited by a moderator: Mar 23, 2011