I ran a study with the following design: Subjects were presented 100 different stimuli and asked to indicate their liking (scaled from 0-10) for each stimulus. Each stimulus was part of only one of 4 different groups, thus it was a nested design (repeated measures for the different stimulus groups; within-subjects). Now I would like to conduct a linear mixed model analysis using the lmer function of the lme4 package in R, but as a newbie I am very insecure about my approach. I want to consider both random intercepts for stimuli and subjects and also allow random slopes for subjects.

I thought of the following model:

model = lmer(liking ~ group + (1|group/stimulus) + (1+group|subject), data=mydata)

Does this model make any sense?

  • $\begingroup$ Since your response is bounded, wouldn't it make more sense to use beta regression of score/10? $\endgroup$ – Frans Rodenburg Oct 5 '18 at 11:18

Indeed, as Frans Rodenburg suggested, given that you have a bounded outcome you could fit a Beta mixed effects model. In addition, I think that you could start with a random intercept for subject and a random intercept for stimulus, i.e.,

liking ~ group + (1 | stimulus) + (1 | subject)

Such a model can be fitted with the glmmTMB package.


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.