More Mixed Models

In the R world, lme4 is a great package for mixed model estimation, and the most widely used for such models. For standard settings, few tools will do the trick as easily or as quickly, and because of that, its approach has been emulated in other packages and statistical programs. However, that ease and efficiency comes at a price of being able to do more complex models, so at some point you may need to switch gears. This workshop will demonstrate other ways to potentially get what you need.

Demonstration will (potentially) include the following, along with some discussion of strengths and/or drawbacks to use.

• glmmTMB: heterogeneous variances, auto-correlated residuals, zero-inflated models
• rstanarm, brms: Bayesian approaches
• mgcv: additive effects, robust models, big data
• statsmodels: in case you’re in the Python world.

This workshop assumes basic familiarity with mixed models, and familiarity with R, especially the lme4 package. However, this is a brief demonstration with a simple goal of bringing about awareness of other useful tools, so one can simply attend to hear about them.

Click to download this repo. For the workshop, you can run the code in mmm_notebook.Rmd as we go along. If you just want to follow along/see the content, open mmm.html in your browser after downloading the repo.