UIUC Quant Brownbag

When you should and shouldn’t use mixed models and/or Bayes factors

Henrik Singmann - University College London

Two important statistical developments of the last years are relatively easy to use software tools for mixed-effects models and Bayes factors. Because of these tools and pressure from reviewers and editors, many researchers feel the need to incorporate these statistical techniques into their methodological arsenal. In my talk I want to provide a conceptual overview of these methods and provide some recommendation when and when not to use these tools. In short, I recommend that mixed models should only be used when they provide a clear benefit over well-established procedures such as repeated-measures ANOVA (e.g., for crossed-random effect designs or designs with partially crossed factors). I also caution against the use of so-called default Bayes factor (as implemented in e.g., JASP or the BayesFactor R package) for designs involving repeated measures.