I'm very new to R and haven't explored repeated measures before. I have three groups (morphs: br, ye, gr), and multiple response variables that I measured over time (0hr, 1hr and 24hr). My sample sizes are low, with just 3 samples per morph per time point. The samples of the different morphs were taken from 9 individuals (col).

According to leveneTest(rfp~factor(time*morph),data=data), variance is homogeneous for each variable, but non-normal distributed looking at qqnorm for all data points per response variable (I don't know if I can and should look at this for time*morph too, though sample sizes will probably prevent this).

I would like to statistically assess differences in response variables over time within and among morphs. I believe I need to use a GLMM with repeated measures for this, but I keep going round in circles with it. I don't know how to determine which model to use, and then how to use it!


1 Answer 1


Well, I use mixed modeling in R quite a bit. The first thing you should decide is what package to use. The four most common (in my opinion) are: lme4, nlme, saberR, and mcmglmm. The first two are the older and more established packages while the other two are more new. The main difference, besides coding syntax, is that lme4 and nlme are univariate only while saberR and mcmglmm can perform multivariate mixed models.

That said, a mixed model, though, might not be appropriate with only 27 samples in total. One can most certainly be fit, but you might get quite a bit more mileage from something like a non-parametric anova. Personally, I feel that doing a kruskal-wallis H test will be quite a bit more useful.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.