Subscribe to DSC Newsletter

 As we know in insurance sector claims follow exponential fmily distributions, so in oredr to build the claim severity models we use the GLM technique called gamma model.


I am trying to build insurance claim severity model in SAS using proc genmod with gamma distribution. i am getting very large values for 'scale deviance' and 'pearson scale deviance' (i.e value/df) in goodness of fit tests, where in for good model these values should be near to 1 .


what would be the probable reason for this and how could i get better results and is there any literature on gamma models used for claims severity modeling to cross validate the procedure. 



Tags: claims, gamma, insurance, model

Views: 1458

Reply to This

Replies to This Discussion

Observe the histogram for claim severity and verify whether the model can be used. Or provide more information about mean, median and standard deviation of claim severity.

Sir, i have checked histogram, it is showing gamma distribution gamma(2,3) shape, with the following details
mean=1972, median=900, sd=3621
It seems that the distribution is positively skewed; seen from details given by you. Then use the data to fit a Gamma Distribution using say Method of moments or MLE. Once getting the parameters, they may give some information and the associated Chisquare goodness of fit test will ensure whether it could be statistically valid. Please write if you need further help to fit the distribution.

Check the value against chi-square to see if the results are reasonable.  If not, take a closer look at the residuals to see if you can pinpoint what is going on.


-Ralph Winters


On Data Science Central

© 2021   TechTarget, Inc.   Powered by

Badges  |  Report an Issue  |  Privacy Policy  |  Terms of Service