Excel Problem: GLM_ModelRefinement1

I may be missing something obvious, but can you explain where the $45,000 comes from in the solution for part a? How did we know it was the upper range of Vehicle Value?

Comments

  • Good catch, I accidentally left the solution at $45,000 instead of setting it back to $300,000 after sensitivity testing.

    You should have ln(u/(1-u)) = 61.063 which gives the probability of a claim as 100%.

    From our answer in part a we should question whether predicting a claim will almost certainly occur is reasonable. At this point we can then use trial and error to figure out where the model stops being reasonable. This is where the $45,000 figure comes in. What likely happened was the model was fit on a lot of low value vehicles and then a continuous linear fit found. The resulting end-scale behavior is poor. So the model needs refining by using a higher order polynomial for vehicle value, breaking vehicle value into categories, or adding additional rating variables.

  • Appreciate the clarification and extra rationale!

Sign In or Register to comment.