I was wondering: I have read that people use lsmeans() and glht() to perform multiple comparisons of a lmer() model. I have the problem that it does not work with my dataset, as the two models tell me that there is no contrast detected (see more here: https://stackoverflow.com/questions/48241097/glht-and-lsmeans-cant-find-contrast-in-lmer-model).

Generally, why can I not use multiple lmer() to make multiple comparisons? Technically, if I perform a lmer() on 4 levels, given the same sample-size, should it be not equal to perform a lmer() on 2 of those levels only?

  • $\begingroup$ Honestly they are not the same question. The one on Stack Overflow is about a problem with the software, this is about the method. And I have separated them to 1) make them useful to people that in future will be concerned with either this or the other question; 2) to avoid putting a question about method in a website concerned with computer programming concerns. For example, to this question, another student has answered: "As far as i can tell random effects models if correctly specified are nearly always more powerful than linear models, so if anything you should... $\endgroup$ – Luca Danieli Jan 14 '18 at 14:26
  • $\begingroup$ ... prefer an lmer with smaller sample size".. Please, don't think that things are so obvious $\endgroup$ – Luca Danieli Jan 14 '18 at 14:27

Your Answer

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

Browse other questions tagged or ask your own question.