Hi for all, I am trying to fit NB2 model for taxa from an open-source, phyloseq object. I analyzed the taxa carefully, plotting histograms with log10 scale or clr. I tried to check the GOF tests for taxa from different count models, i.e. Poisson, quasi-poisson and NB2. I used AIC, BIC and Pearson-chi-squared. However, NB2 is doing better than other models, but GOF tests still show poor results, for example, AIC = 9547 for Akkermansia with a degree of freedom = 2. I have used DESeq2 many times before, I think it is an awesome method, but I can not figure out these results. I think I am familiar with DESeq2 tutorial also but still, I am not an expert. Any suggestions or comments are highly appreciated.

I believe this is a phyloseq question perhaps? We don’t have GOF tests for various count models.