Just to confirm,
I see the negative binomial distribution is used in brms (in my case) is
zero_inflated_neg_binomial_log
If I want to input a dispersion from gene expression testing methods (e.g. edgeR)
This histogram shows the dispersion estimation from an external model for 20K genes.
which of these is correct?
shape ~ 1 + offset(log(1/edgeR_overdispersion))
shape ~ 1 + offset(1/edgeR_overdispersion)
shape ~ 1 + 1/edgeR_overdispersion