10-30-2014 09:27 AM
I work in insurance as a pricing actuary and use SAS extensively for the purpose of preparing claims and exposure data that can be used to carry out GLIMs in a seperate software - Emblem by Tower Watson.
I have heard that it is possible to perform GLIMs in SAS but I am trying to find a useful source for the code to achieve this. Does anyone know if:
1. It is indeed possible to perform a GLIM in SAS
2. Is there code publicly available that will perform same
To illustrate what i am looking to do I will explain what I am modelling. I work in commercial fleet insurance where I want to model the claims experience (claims frequency using Poisson error distribution and claims severity using Gamma and a log link function) for each of the main claims types i.e. Third Party Damage, Third Party Bodily Injury, Windscreen etc. and the idea is that I model the effect that various rating factors have on the reponse variable. The main rating factors I use are Vehcile Type, Cover Type, Excess, Industry, Location, Calendar Year and Excess.
In Emblem (an expensive software), when you import the dataset for a particular model, say Windscreen Claim Frequency, you just select the error distribution you want (Poisson) and all factors in your model are on the left of the screen and a graph showing the actual observed, model curve etc, is shown on the centre of the screen. It's very easy to use and very flexible but it doesn feel like you are operating a blank box and I would like to do this in SAS if possible to undertand the inner workings of the model in more detail.
Many thanks in advance
10-30-2014 03:48 PM
I would expect that PROC GENMOD is the closest procedure. I would recommend looking through the documentation and examples to see if it matches what you want to do.
10-31-2014 11:54 AM
Ditto what Steve said...just adding that the documentation is somewhat massive in scale, but one of the introductory examples is exactly a Poisson claim frequency model: