I'm working with GLMM and a binomial distribution to find the best model for my biological data. I'm currently writing my PhD and I have some trouble to present my results. I read and learn a lot about GLMM but I don't find a satisfying way to present my results, as I'm not statistician.

I'm using R and the lme4 package. I have different GLMM models, with different random intercepts. I compare them with a likelihood ratio test to find the smaller AIC and check if my variables were significant (by removing them one by one).

I did also cross-validation to show how the different models predicts on "unknown" dataset. My questions are pragmatic:

-> should I display all the models with the AIC values and LRT?

-> should I run a cross-validation for all of them?

-> from the cross-validation, is it ok to only present the result from the AUC roc curves?

-> which other results/parameters would you expect to see?

I know that my questions might be terribly trivial but I need some answers as simple as possible as I'm completely lost.

  • $\begingroup$ Please indicate your study objectives and show part of your data compilation. $\endgroup$
    – user10619
    May 7, 2019 at 8:31

1 Answer 1


You have not told us which specifically are your aims with these models. Note that statistical models, in general, are most often used either for effect estimation (i.e., to see how specific factors/covariates are associated with your outcome) or prediction (i.e., which terms you want to include in your model to obtain the most accurate predictions). Depending on the aim you have different model building strategies apply. For example, for effect estimation most often you do not need to remove “non-significant” variables from your model.

  • $\begingroup$ The aim was to find a model which give me the best predictions of a binary dependent variable. $\endgroup$
    – RforLife
    May 5, 2019 at 20:14

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.