Topics

Bayesian multilevel modelling workshop 2021

Topics

Planned contents

This workshop is designed to provide a practical introduction to basic and advanced multilevel models. Participants will learn how to fit models using both maximum likelihood and Bayesian methods, although the focus will be on Bayesian parameter estimation and model comparison. We will start with a short introduction to multilevel modelling and to Bayesian statistics in general, followed by an introduction to Stan, a probabilistic programming language for fitting Bayesian models. We will then learn how to use the R package brms, which provides a user-friendly interface to Stan. The package supports a wide range of response distributions and modelling options, and allows us to fit multilevel generalized linear models. Depending on participants’ wishes, we will take a closer look at modelling various types of data, such as choices, response times, ordinal or longitudinal data.

Specific topics include:

• Bayesian inference: an introduction

• Bayesian parameter estimation

• Model comparison & hypothesis testing

  • Bayes factors
  • Out-of-sample predictive accuracy (LOO)

• Specifying multilevel generalized linear models

• Understanding statistical models through data simulation

• A principled Bayesian workflow for data analysis

Your expectations/questions

Since most of you expressed an interest in Bayesian statistics, we will mostly multilevel generalized regression models from this perspective.

Specific topics

• advantages of Bayesian over frequentist statistics
• how to select priors
• preparing data
• binary (choice) data
• response times
• repeated measures and other hierarchical (multilevel) designs, inlcuding longitudinal data
• crossed random effects, e.g. lexical decision tasks
• categorical variables
• moderation / mediation
• SEM (structural equation models)
• model comparison / hypothesis testing in mixed effects models
• dyadic data
• nonlinear regression