Comparing overdispersion distributions

Asked Sep 14 '12 at 12:33

Active Sep 18 '17 at 20:50

Viewed 406 times

I am dealing with handling overdispersed count data (Poisson distributions fails to fit).

I need to compare three different mixtures (Gamma, Log-Normal and Inverse Gaussian) of Poisson rate parameter distributions using a simulation study.

I would like to know the available methods (apart from Likelihood ratio and Pearson's Chi Square test) to compare three completely different distributions to fit overdispersed count data.

Any reference also highly appreciated!!

edited Sep 18 '17 at 20:50

kjetil b halvorsen

63,378
26
142
467

asked Sep 14 '12 at 12:33

Arun

Why wouldn't you also consider a negative binomial model? – Michael R. Chernick Sep 14 '12 at 16:20
As for Michael's comment, note that Poisson with parameter gamma distributed __is__ the negative binomial model .... It could be easier, speaking of programming, to use the negative binomial directly. – kjetil b halvorsen Sep 15 '12 at 01:35
You could trey a qqplot against the three distributions – kjetil b halvorsen Jan 17 '18 at 22:51

Comparing overdispersion distributions

0 Answers0

Linked