I have percentage areas values from 4 treatments (25 replicates per treatment). I would like to compare these percentages.
I am supposed to use a beta regression, because my response variable is a proportion (not resulting from a count). For other variables I use GLMM and then use the glht() function to get the significance of pairwise comparisons. This does not seem to work with my betareg model. Is there any way to do that ?
Also I would like to include random effects because of nested sampling protocol. From this answer this does not seem very appropriate/easy to do this with beta regression.
Do you have any suggestion ?
Thanks to @rvl's comment I could calculate pairwise comparisons and extract group letters.
library(lsmeans) library(betareg) betalive = betareg(Live ~ Crop) live.rg <- ref.grid(betalive) live.lsm = lsmeans(live.rg, "Crop") cld(live.lsm, alpha = 0.05)$.group >  " 1 " " 1 " " 1 " " 2"
However I could not solve my problem of random effects yet.