Any thoughts on this paper? I thought it was ok - nothing too suprising or nasty in there (unless I missed something), and quite a few easy marks Btw, was the answer to the ILF question (question 2)curve C?
Agree nothing surprising in here, and questions very broken up which was helpful. Curve C is right (I think!).
I was a bit unsure what they were looking for with the 'why would the long term average cat losses experienced differ from the cat model expected losses, other than due to random variation' question. I took a bit of a scatter gun approach - not sure if they were looking for reasons specific to cat models, or mostly more general pricing type things, data related things, or reasons for different experience between insurers etc. Also, on the last question, on GLMs, I'm not sure what 'definition of a GLM' they were looking for in the first part - I guessed they wanted a formula with link function, explanatory variables etc, but there were several different forms of these formulae, so i'm not sure if they had something else in mind.
Yes, the cat question was a bit harder. Scattergun is a good approach because you are sure hit a few of the marks here and there! I guess model errors of various kinds could be relevant for this one, as well as simulation error and timing differences. And I wrote out the standard glm formula, defining the parts - link, design matrix, parameters, fixed effects, error term.
