Introduction to Bayesian statistical modelling

Level

PhD students and postdocs of the Faculty of Psychology and Educational Sciences of Ghent University

Organizing Committee

Used Software

The course will be taught using R (free statistical open-source software) and the RStudio GUI, and packages relevant for Bayesian data analysis (mostly the brms package).

Content

Day 1 will start with a general overview of the model comparison approach to data analysis (as implemented in R), to guide participants from a “hypothesis testing” to a “parameter estimation” mindset. Notions of model comparisons will subsequently be outlined, in order to provide practical ways to implement model selection methods in everyday research. The Bayesian data analysis framework will then be introduced, with particular emphasis on how formalized prior knowledge and observed data are integrated to generate posterior distributions.
Day 2 will be devoted to the introduction of the brms package by its creator, Dr. Paul-Christian Bürkner. Practical examples will follow, focussing on data analysis problems typically encountered in everyday research: modelling of continuous data via linear regression, generalized linear models for categorical (count and percentage) data, and ordinal regression to adequately analyse data from Likert scales.
During day 3, Dr. Bürkner will introduce multilevel models and its straightforward implementation in brms. Examples of models built during day 2 will be extended to estimate uncertainty arising from subject-specific variability. Finally, Dr. Bürkner will show how to conduct meta-analyses within a hierarchical Bayesian framework.
Participants will be technically supported by the course organizers (Grahek, Nalborczyk, & Schettino) throughout the whole duration of the course, and will additionally be encouraged to ask questions regarding their own data and research plans.

Objectives

Students start by acquiring basic skills in order to be able to understand and conduct Bayesian data analysis in the R software environment. This allows students to understand the analyses and complete the hands-on exercises during the second and third day of the course. At the end of this course the students should:

  • Be familiar with the R environment and the available packages for Bayesian data analysis
  • Know the elementary commands and statements used in R and Bayesian data analysis packages
  • Understand the theoretical underpinnings of Bayesian statistical models and their potential applications
  • Be able to report output of Bayesian data analysis

Instructor

  • Paul-Christian Bürkner, Research Associate, Department of Psychology, University of Muenster WWU

Fliednerstr. 21
48149 Muenster, Germany
Tel: 0251 83-39418

https://research.aalto.fi/portal/paul-christian.burkner.html

  • Paul-Christian Bürkner will be the main lecturer of this course. Paul is a research associate in the Psychology department at the University of Muenster. His PhD dissertation, completed in 2017, was awarded the Gustav A. Lienert award for the best methodological dissertation. During his PhD, he developed the brms package (Bürkner, 2017), currently one of he most used R packages for Bayesian data analysis. Over the past years, Paul has published several peer-reviewed manuscripts regarding Bayesian data analysis and meta-analysis. Furthermore, Paul has been invited to give workshops on Bayesian data analysis at several universities. He also has extensive experience as a lecturer in statistic methods, for which he received this year the award of the best lecture of the Institute of Psychology in Muenster. Given his expertise on the topic and his innovative work, Paul will lead the lectures during day 2 and day 3.

Study Material

The instructors will provide the students of a syllabus and will give theoretical presentations which will be alternated with practical demonstrations and hands-on computer exercises in the R environment.

Format

Theoretical lectures will introduce basic concepts, followed by demonstrations of analytical procedures and practical sessions to get hands-on experience. During the second and third day two lecturers will be present to guide the practical exercises.

Dates and Venue

  • 13-14-15 June 2018 (Wed to Fri)

Room 1.3 (PC-Lokaal 1), Faculty of Psychology and Educational Sciences, Henri-Dunantlaan 2, Ghent University, 9000 Ghent

Programme

  • Day 1 - Wednesday 13 June 2018: Statistical modelling, model comparison and Bayesian statistics (Ivan Grahek, Ladislas Nalborczyk, & Antonio Schettino)

9:00 Registration
9:30 Introducing the model comparison approach to data analysis
12:00 Break
13:00 Notions of model comparison (information criteria, Bayes Factors)
15:00 Introduction to Bayesian data analysis
17:30 End of day

  • Day 2 - Thursday 14 June 2018: Introduction to Bayesian statistical modelling (Paul-Christian Bürkner)

9:00 Introducing the brms package, a simple linear regression case study
12:00 Break
13:00 Toward generalised linear models, modelling categorical data
15:00 Ordinal regression with brms, modelling Likert-type data
17:00 End of day

  • Day 3 - Friday 15 June 2018: Advanced Bayesian statistical modelling with brms (Paul-Christian Bürkner)

9:00 Introducing multilevel models with brms
12:00 Break
13:00 Toward generalised multilevel models
15:00 Meta-analysis with brms
17:00 End of day

Registration and information

Please follow this link: https://webappsx.ugent.be/eventManager/events/brmsUGent

Registration fee

Free of charge for PhD students of the Doctoral School of Social and Behavioural Sciences of Ghent University.

Participants

Maximum 20 participants

Evaluation criteria (doctoral training programme)

Presence and active participation during all sessions