APPLIED INTRODUCTION TO BAYESIAN DATA ANALYSIS
Copenhagen Graduate School of Social Sciences
January 5-7, 2015

Please note: This page is archived from the last time I taught this course.
For materials for current courses, please go to my teaching page.

Welcome to the website for my 3-day workshop on Bayesian Data Analysis at the University of Copenhagen, January 5-7, 2015. This page contains materials for workshop participants and additional resources for Bayesian data analysis.

Workshop outline

The outline for this workshop is available here.

Setup & before we meet

Before our first meeting, please complete a couple of steps to prepare for the workshop:

1. Complete my brief survey to give me a rough idea of the background of all workshop participants. Please email your answers by January 1.
2. Have a look at my "(very basic) R tutorial" and install R and RStudio on your computer.
3. Have a look at sections 1-3 of my tutorial "Using JAGS via R" and install JAGS and the packages mentioned in section 3 on your computer.

We will spend the last part of the workshop on Monday to work through the remainder of the "Using JAGS via R" tutorial. Please bring your laptop computers (if available) to the workshop. If you run into problems installing any of the software packages mentioned above, I'll be happy to assist you during the first day of the workshop.

Tutorials

The tutorials below contain information and code to .

1. A (very basic) R tutorial
2. Using JAGS via R
3. Presenting Bayesian model output

Slides and source code

Slides for each workshop unit and the source code I used to create them are posted below. Slides are posted as .pdf files, and the R code for all applications is posted in .R files.

• Unit 1
• Unit 2
• Unit 3
• Unit 4
• Unit 5
• Unit 6
• Unit 7
• Unit 8
• Unit 9

Modeling exercises

This file provides some (voluntary) exercises for all topics discussed in the workshop.

Modeling examples

The R scripts below contain fully working examples of Bayesian regression models for different types of outcome variables. All models are fit using JAGS.

• angell.instructions.R - fit a Bayesian linear regression model, using the Angell data from John Fox's R and S-PLUS Companion to Applied Regression in R using R2jags. This example contains code for data preparation, model fitting, and diagnostics.
• angell.jags.zip - fit the same model directly in JAGS from the Terminal and analyze the output using the coda and boa packages in R.
• beer.instructions.R - fit a Bayesian ordered logit model, using the Beer data in R using R2jags
• econvote.instructions.R - fit a Bayesian multinomial logit model, using Austrian voting data in R using R2jags. [Multinomial logit model]
• undervote.instructions.R - fit Simon Jackman's Undervote example in R using R2jags [Difference of two proportions]
• turnout.instructions.R - fit (a slightly modified version of) Simon Jackman's Turnout example in R using R2jags. [Logit model]
• legislators.instructions.R - fit Simon Jackman's Legislators example in R using R2jags. [Bayesian IRT model]
• pumps.instructions.R - fit the Pumps demo from the WinBUGS examples folder in R using R2jags [Conjugate gamma-Poisson hierarchical model]
• rats.instructions.R - fit the Rats demo from the WinBUGS examples folder in R using R2jags [Normal hierarchical model]
• wvs.instructions.R Linear model with group-level coefficients using an old subsample of World Values Survey.
• mlm.state.instructions.R - Step-by-step instructions to set up hierarchical data in R and fit a multilevel logit model using R2Jags. Data: polls.subset.merged.dta, polls.subset.JK.dta, presvote.dta.

Diagnostics

The bottom of this script shows a variety of ways to obtain diagnostics (traceplots, density plots, BGR, etc.) of JAGS/MCMC objects in R using the coda, superdiag, ggplotmcmc, and mcmcplots packages.

Model presentation

• regression.table.R - R code to easily export JAGS/BUGS results to LaTeX or HTML.
• regression.dotplot.R - R code to easily make regression coefficient plots from JAGS/BUGS results.
• interaction.instructions.R - R code to plot marginal effects from a Bayesian linear model with an interaction term across the range of a moderating variable.
• logit.pp.plot.instructions.R - R code to plot predicted probabilities from a Bayesian logit model over a simulated range of values for an explanatory variable of interest.
• ologit.pp.plot.instructions.R - R code to plot predicted probabilities over the simulated range of explanatory variables after a Bayesian ordered (or multinomial) logit model. Partially based on code by Dave Armstrong (UW-Milwaukee).
• factor.dotplot.R - R code to make a dot plot (with credible intervals) of a Bayesian factor score.

Questions? E-mail me.