LMM for repeated-measures I want to perform a repeated measures analysis on an unbalanced dataset. To keep it simple, assuming I have activity level (AL) as my response variable, time  as my repeated measures. 
Additional factor don't matter for now. 
My time variable is not continuous, but I measured for 3 different stages during an intervention. My goal is to draw conclusions about changes in AL from one stage to another. As far as I understand, the LMM would tell me if 'time' has an influence on the AL in general. But I need to know whether there is a significant change between any of my 3 single time points (e.g. between stage 1 and stage 2 etc..). What would be the correct approach to do so? Do I need to perform multiple-comparisons for each stage?
 A: Yes, you can use a linear mixed model and then perform multiple comparisons. One way to do this in R, with lmer from the lme4 package and glht from the lsmeans package, is:
> require(lme4)
> m0 <- lmer(effort ~ 1 + Type + (1|Subject), data=ergoStool, REML = 0)
> summary(m0)

Fixed Effects:
(Intercept)       TypeT2       TypeT3       TypeT4  
     8.5556       3.8889       2.2222       0.6667  

Here the estimates for each level of Type are contrasts with the baseline level (T1). To get pairwise comparisons between other levels we can use:
> require(lsmeans)
> summary(glht(m0, lsm(pairwise ~ Type)))

which outputs the desired contrasts:
Linear Hypotheses:
             Estimate Std. Error t value Pr(>|t|)    
T1 - T2 == 0  -3.8889     0.4890  -7.952  < 0.001 ***
T1 - T3 == 0  -2.2222     0.4890  -4.544  < 0.001 ***
T1 - T4 == 0  -0.6667     0.4890  -1.363  0.53140    
T2 - T3 == 0   1.6667     0.4890   3.408  0.00956 ** 
T2 - T4 == 0   3.2222     0.4890   6.589  < 0.001 ***
T3 - T4 == 0   1.5556     0.4890   3.181  0.01689 *  

