topics.Rmd

---
title: "Topics"
description: | 
  Bayesian multilevel modelling workshop 2021
date: 05-21-2021
author:
  - first_name: "Andrew"
    last_name: "Ellis"
    url: https://github.com/awellis
    affiliation: Kognitive Psychologie, Wahrnehmung und Methodenlehre, Universität Bern 
    affiliation_url: https://www.kog.psy.unibe.ch
    orcid_id: 0000-0002-2788-936X

citation_url: https://awellis.github.io/learnmultilevelmodels/topics.html
bibliography: bibliography.bib
output: 
    distill::distill_article:
      toc: true
      toc_float: true
      toc_depth: 2
      code_folding: false
---

```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = FALSE)
```

# 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