Hi
I have the following puzzle

Several places in the (also the pdf) help for -pwcompare-, option mcompare it says: "+ tukey, snk, duncan, and dunnett are only allowed with results from anova, manova, regress, and mvreg."
And looking into formulas I think it is because the same variance is assumed for all observations.
The puzzle is that R apparently (I'm no expert in R, but so I'm told) has Tukey method as default using emmeans after lme.

Does anyone know of references saying why it is either possible or impossible to use the Tukey method after mixed regression?