Bayesian Data Analysis

or: Practical Data Analysis with BUGS using R

A short course taught by Lyle Gurrin
Monday 13 - Friday 17 August 2012, Copenhagen

Bayesian inference has become increasingly important during the last decade. This is largely attributable to the computational possibilities that have emerged, notably the easy access to MCMC methods. Lyle Gurrin has developed a highly acclaimed course in Bayesian data analysis for the Biostatistics Collaboration of Australia which he has previously presented in Stockholm and Copenhagen. The course is based on the textbook Bayesian Data Analysis by Gelman et al. The course, like the book, emphasises practice over theory.

The course will be run over 5 days, with alternating lectures and practical computing sessions. We will work through several case studies, each comprising a lecture and a laboratory session. Example analyses will include meta-analysis of clinical trials, longitudinal analysis and method comparison studies. Throughout the course we will illustrate differences between classical and Bayesian approaches to analysis of the same problems.

The practical part of the course will be based on the language BUGS for model specification, but we will primarily be using the JAGS implementation through the rjags package for R, because this works the same way on Windows, Mac and Linux.

The emphasis in the course will be on illustrating how to take the practical examples all the way from data formatting to the reporting stage in terms of tabulations, graphics etc. Participants are requested to bring their own laptops, we will provide instructions in advance about what to install before the course.

Target Audience

The intended audience is biostatisticians, epidemiologists, and public health scientists wishing to gain insight and practical experience with Bayesian data analysis. The material in lectures is aimed at students of biostatistics and covers the statistical theory at an undergraduate level. The practical exercises will be fully scripted and directed with three tutors on hand, so they should be manageable also by those who are not experts in statistical theory and programming.
The course will be limited to 25 participants.


Last modified: 1 January 2013, BxC