I'm currently working with the glht function from the "multcomp" package, from R, to run a Tukey's post-hoc test (after a LMM analysis). I found two scripts to do that: one considers only Tukey's (e.g., summary(glht(MY MODEL,linfct=mcp(MY FIXED FACTOR="Tukey"))) ), and the other uses Holm's test as an adjustment (e.g., summary(glht(MY MODEL,linfct=mcp(MY FIXED FACTOR="Tukey")), test = adjusted ("holm")) ). My question is about the statistical power of the models. Is one model more conservative than the other? Is one of them better in terms of data confiability?