Bayesian Hierarchical Modeling with R and OpenBUGS/JAGS

Course title

Instructor

Dr. Pablo E. Verde. Senior Researcher at Coordination Centre for Clinical Trials. Heinrich-Heine-Universität Düseldorf, Germany.

Course language

English

Course schedule

July 6 - 10 from 10:00am to 1:00pm

Description

Data scientists and researchers usually need to analyze data with a hierarchical or multi-level structure, missing values, imprecise measurement data, complex correlation patterns and other complexities found in practice.

Bayesian hierarchical models offer a natural approach to handling these types of problems by the construction of statistical models which reflect the complexity of the data.

In this course we make emphasis in visualizing and exploring hierarchical data by using powerful graphical tools in R. The model building is performed by using Bayesian graphical models and computations with Markov chain Monte Carlo methods by linking OpenBUGS (or JAGS) software with R.

Program

Day 1: Introduction to Bayesian data analysis and R/BUGS

Introduction to Bayesian analysis in practice
Getting started with R/OpenBUGS/JAGS
Practical 1

Bayesian graphical models and MCMC computations
Bayesian regression and multiple parameters models
Practical 2

Day 2: Bayesian Hierarchical Modeling with R/BUGS

Visualization hierarchical data with R
Statistical framework for Bayesian hierarchical models
Practical 3

Prior distributions for hierarchical models
Bayesian hierarchical models for longitudinal data
Practical 4

Day 3: Advance topics in Bayesian Hierarchical Modeling

Predictions, model checking and model comparison for hierarchical models
Multilevel modeling, complex patters of variation and other extensions
Practical 5

Hierarchical models for missing data problems
Bayesian evidence synthesis and meta-analysis
Practical 6

Target audience

Participants should be familiar in classical data analysis, ideally they should have some experience in applying mixed effects models (e.g. by using R, SPSS or SAS).

To attend the course you do not need experience with Bayesian analysis or with OpenBUGS, these topics are covered during the course. It is a great advantage to be familiar with the statistical software R.

The course presentation is practical and not theoretical with many worked examples. Lectures are given in English or in Spanish.

Evaluation

After each topic participants have to make practical applications of the Lectures’ material. These exercises are taken as evaluation of the course.

Classroom

PC3