Overdispersed Counts

[tmeeha]

Jonas, this is a great package! Community, I have a few questions about using the package.

1. I am working with overdispersed count data. Is it possible to specify a negative binomial distribution for the response?

2. If not, one solution to dealing with overdispersion is to have an observation level random intercept. If my simple model is:
   model1 = list(
   count ~ 1 + year_c,
   ~ 0 + year_c
   ),
   can you give me a hint as to how to add an observation level random effect? I have a small dataset (n=20) so hopefully this isn't too computationally challenging.

3. I am looking forward to v0.04 when we can add a global covariate to the formula list. Right now I get an error when I add a second covariate to any part.

Thanks again Jonas for the great package and to the community for help.

Tim

[lindeloev]

v0.5 should include custom families and probably a larger set af default families (v0.4 will include most of the infrastructure work to enable it). But as things stand now, I can't give an ETA since I prioritize another project higher.

One solution is to look at the fit$jags_code and make the changes necessary to do NB and/or random effects. But all the mcpfit-specific convenience functions of mcp (plotting, summaries, etc.) would probably not work out-of-the-box with these.