TPI frequency model

Hi



I'm looking at a TPI frequency model for fleet and I wanted to check something.



Once of the factors I have is vehicle value which is very significant in the model - it brings the deviance from 17,000 down to 16,000 whereas a lot of the significant factors lower the deviance by only a fraction of this.



I fitted the factor as a Variate with 3 polynomials to capture the curvature.



The only issue now is that loads of my other factors are highly insignificant where the levels have huge standard deviations (1000%). Does this just mean that vehicle age is just correlated with loads of factors and is sucking out any effect from a lot of other factors or should I be cautious about this and do some investigating. When looking at private motor I don't remember getting such dramatic results. The other factors that are highly significant are failing the chi squared test horrendously (i.e. 70% to 100%). It just seems a bit odd as I was expecting the other factors to be more significant than this.



The factors I'm talking about are:

• area


• no_of_risky_drivers


• age_of_youngest_driver


• vehicle_body_type


• vehicle_make


• vehicle_type

Perhaps this is just a feature of the claim type TPI?

TPI frequency model