Linear Mixed Models Resources

Next month, I’ll be travelling back to Queen’s University Belfast, my old department, to give a two-day workshop on linear mixed models. I spent some time this morning putting together some pre-workshop material for attendees. Afterwards I realised some of these links might come in useful for others at some point along the way, so here they are!

[…]

The workshop will be based around a handful of commonly used resources.

We’ll be using the statistical language R (https://www.r-project.org/), with the RStudio editor (https://www.rstudio.com/) throughout. In R, we’ll mostly be using the dplyr package for data manipulation, ggplot2 for visualisation, and lme4 for model fitting. Most of the packages needed are part of the ‘tidyverse’ (https://www.tidyverse.org/) which, I suggest you check out before the workshop. If time allows, we’ll also look at using the brms package (https://github.com/paul-buerkner/brms) to fit Bayesian multilevel models. brms can be a little tricky to install on Windows computers, and Bayesian models are slow to fit, so we won’t try to actually fit these models during the workshop. However, this is one of the most powerful statistical tools to come out in recent years, and is well worth checking out.

I don’t know yet whether computers will be provided, or if it will be necessary to bring your own. In any case, you’ll want to install R, RStudio, and the packages mentioned above on your own computer. All of these tools are free and open source, and can be installed on university computers without administrative access.

Most of the workshop content will be based around the book Advanced Regression Modelling, by Andrew Gelman and Jennifer Hill (ARM; http://www.stat.columbia.edu/~gelman/arm/). This is a fairly substantial text, and is one of the most useful statistics resources around. It’s also possible to find electronic copies of this book through the library, or through Google. ARM is a little old, but it is possible to find draft content from the updated version, Regression and Other Stories, at http://www.stat.columbia.edu/~gelman/regression/

ARM is not aimed specifically at psychologists or cognitive scientists, although it does include a few good examples from our field. For this reason, I’ll be going through some additional material from my own experience, or from other sources. Jake Westfall has saved me quite a lot of effort by putting together a list of some of these resources back in 2015, at http://jakewestfall.org/blog/index.php/2015/06/20/reading-list-introduction-to-linear-mixed-models-for-cognitive-scientists/. Some of these resources are a little out of date, but it still makes for a useful starting point. Westfall’s blog itself is also excellent, and worth reading.

Other useful resources include

Also