Gamma Mixed Model, Posthoc Odds Ratio

Discuss statistics related things
Post Reply
Posts: 6
Joined: Thu Jun 24, 2021 4:34 pm

Gamma Mixed Model, Posthoc Odds Ratio

Post by bencturnbull »

Hi Friends,

When I run a generalized linear mixed model, with a gamma distribution and log link, I get no exp(B) values for the posthocs. I do, however, get them for the fixed effects parameters. I also get them if running a negative binomial distribution.

Is there a solution?
Is there just no way to estimate the effects size of posthocs from a gamma distribution?

User avatar
Posts: 459
Joined: Thu Mar 23, 2017 9:24 pm

Re: Gamma Mixed Model, Posthoc Odds Ratio

Post by »


GAMLj3 implements posthoc tests for beta models (following R emmeans package) testing differences in probabilities (o proportions). So, the posthoc values are basically marginal effects for categorical independent factors. As such, the exp(b) would not make much sense. However, you can always get the effect size of the comparison from the "Parameter Estimates" and use the adjusted z-tests and p-values from the post-hoc table.
Post Reply