# GLM model selection using AICc with Tweedie distribution

I have two questions regarding use of Tweedie GLM in R. I am new in using this distribution and despite a thorough search on different forums, I could not find my answers.

I am now running several GLM models using Tweedie distribution because of compound Poisson-Gamma distribution of my data with a spike around zero.

I can successfully run the models but firstly, AIC is not provided in the model results. There is a command for AIC calculation (AICtweedie) but I could not find anything calculating AICc (as my data points are few). Is this just matter of lack of enough interest in Tweedie that such a command doesn't exist or there is some logical or technical constraints behind it?

Second, use of packages such as MuMin or glmulti is not possible when using Tweedie distribution (because AIC values are not calculated in Tweedie models). Has anyone got any similar packages to run multiple models and rank them based on AIC(c) when using Tweedie? Running around 500 models and type down the AIC values is not easily possible.

• This is R-related and not stats question so I think it should be migrated to Stackoverflow. Mar 15 '15 at 22:00
• @KamilBartoń Arash is looking for a tool to calculate AICc for tweedie models. On SO, this would be closed as off-topic (looking for a tool). Mar 16 '15 at 18:56

You can use AICtweedie directly in MuMIn's functions, just specify it as a rank argument. Alternatively, you could write a wrapper around AICtweedie.
• AICc = AIC+ 2K(K + 1) / (n − K − 1), where K = attr(logLik(model), "df") and n = nobs(model) Mar 17 '15 at 11:13