I have a question about how to remove residual outliers in linear mixed-effects models (i.e., data points with standardized residuals exceeding 2.5 standard deviation units). Should I remove these outliers after fitting the data to the simplest model (i.e., that including only random effects)? Or should I remove them after fitting the data to the full model (i.e, that including fixed effect, interactions and random effects)?

  • 1
    $\begingroup$ Why do you want to remove them? They may be the most interesting finding from your study. $\endgroup$ – mdewey Oct 1 '16 at 16:41
  • $\begingroup$ Because I guess that significance of fixed effects may be biased by them. $\endgroup$ – perper Oct 1 '16 at 16:45
  • $\begingroup$ Have you considered whether a more robust error distribution - e.g. t-distribution may be better? Then a few outliers are less influential. $\endgroup$ – Björn Oct 1 '16 at 16:50

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.