Tutorial on Bayesian LMMs at RHUL 27/3/23
In this short tutorial, I will show one way of fitting Bayesian linear mixed effects models using brms. This repository contains everything you will need to replicate the code in this script. The data comes from a continuous primed lexical decision task, and we will be analysing whether the participants' ERPs in the (pre-defined) N400 spatiotemporal window were more positive when the targets were preceded by semantically related as opposed to unrelated primes (i.e., N400 priming effect). We'll do the analysis step by step:
- make sure the data is as it should be
- contrast coding
- setting the priors
- prior predictive checks
- model fitting & diagnostics
- posterior predictive checks
- sensitivity analysis
I will show how, in this particular example, model fit can be improved by using distributional regression with a random effects structure not just for the location parameter but also for the scale parameter.
If there is time, I will also provide an example of how to analyse response times data with Bayesian LMMs.
NB: Please note that this is not a tutorial on the theory underlying Bayesian data analysis. In this tutorial, I will assume that you are familiar with the fundamental concepts and will only discuss how one can fit models in R using brms.