there are a lot of questions about post-hoc tests for GLMMs on this site and thanks to the replies I almost have my question solved. I want to see if there is a difference in treatment groups over time but for all pairwise comparisons.
My model has a count response (count) with a categorical predictor treatment (as factor with 3 levels: A,B,C) and year (repeated measures of count over time). I also include an interaction to test whether the treatment levels differ over time as follows (simplified here, removed random effects for brevity):
fit <- glmmTMB(count ~ treatment + year + year:treatment)
Using the posts here and here (along with the emmeans vignettes) I have contrasts between groups for each year separately:
emmeans(fit, pairwise ~ treatment | year, at = list(year = c(1, 2, 3,4,5,6)))
$emmeans
year = 1:
treatment emmean SE df lower.CL upper.CL
A 13.55 0.276 423 13.00 14.1
B 13.51 0.276 423 12.96 14.0
C 13.24 0.276 423 12.69 13.8
....
year = 6:
treatment emmean SE df lower.CL upper.CL
A 12.90 0.241 423 12.43 13.4
B 12.89 0.241 423 12.42 13.4
C 12.58 0.241 423 12.11 13.1
But I want all pairwise contrasts for treatment:year to compare treatments within and across all years. This is easy if it were a linear model followed by Tukey's test. The result looks something like:
A:year1 - B:year1
A:year1 - B:year2
A:year1 - B:year3
A:year2 - B:year2
A:year2 - B:year3
etc.
Any help greatly appreaciated.