In relation to the modelling of claim frequency my undertanding is that claims frequency follows a Poisson distribution. Does that mean the errors (observed freq - fitted) automatically follow a Poisson also. So what am asking is if the distribution of the response variable always equals the distribution of the errors?
On a different topic, by taking the log of the dependent variable (in this case claims frequency) is it then linearly related to the explanatory variables?
On a different topic, by taking the log of the dependent variable (in this case claims frequency) is it then linearly related to the explanatory variables?
Error distributions, link functions
0 commentaires:
Enregistrer un commentaire