In the first model, it uses the difference in means and std that are fit to Normal and Uniform distributions, respectively, not the mu and alpha parameters from the Negative Binomial. My question is wouldn’t you want to test the differences after fitting the NB model conditioned on the data? The group1 and group1 NB are unconnected to the BEST. Apologies if I’m misreading / misinterpreting this since I’m new to this application. Thanks.
1 Like