Post-hoc tests in lmer() R


TS Contributor
Hi all,

I'll try to make this my last question of the week. Hard to do this, of course, when everyone is so helpful.

This is a Q for those somewhat familiar with GLMMs in R, as implemented by lmer() in the "lme4" package. I'm well aware of the fact the reported p-values are generally to be avoided. However, I'm stuck with wanting to understand how the levels of one factor differ from one another.

I've examined the glht() post-hoc function in the MASS package. This seems all well and good, but I'm very wary about the multiple comparisons therein, since this is an unbalanced data set.

Any general pointers or workaround? Maybe change the baseline (intercept) in the model to each of level of the focal factor and then use p-values anyways (with Bonferroni). I won't be relying too much on this post-hoc tests for inference, so I'm generally safe. However, preliminary analysis shows interesting patterns.

Any thoughts on post-hocs for GLMMs?