APPLIED BAYESIAN MODELING
ICPSR SUMMER PROGRAM, JUNE 26-28, 2024



Welcome to the website for my workshop Applied Bayesian Models for Social Scientists: From Theory to Estimation and Inference in the 2024 ICPSR Summer Program. This page contains course materials, a variety of information on how to fit and present Bayesian models using Stan and R. Please email me if you have any questions about the material on this page or if you discover any errors. Additional materials for participants will be distributed during the workshop via Canvas. Instructions on how to prepare and what to install will be sent to all participants before the first day of the workshop.

Workshop materials

Installing R, RStudio, and Stan

  1. Install the most recent version of R from the CRAN website.
  2. Download and install the most recent version of RStudio.
  3. To install Stan, a powerful and fast-growing tool for Bayesian estimation, follow the instructions at stan-dev. We've set time aside to go over installation during the workshop, since it will likely require a few individual adjustments depending on your computer's operating system.

Lab tutorials

The following tutorials (available on Canvas) accompany the labs offered during the workshop:
  1. Introduction to R and Reproducible data analysis in R (by Herrissa Lamothe)
  2. Introduction to Stan and rstanarm
  3. Assessing convergence
  4. Working with MCMC output
  5. Managing multilevel data in R
  6. Communicating results from Bayesian analysis

R package: BayesPostEst

To facilitate using the methods from this workshop, I provide an R package (jointly authored with Shana Scogin, Andy Beger, and Rob Williams) to generate and plot postestimation quantities after estimating Bayesian regression models using MCMC. Quantities of interest include predicted probabilities and changes in probabilities in generalized linear models and analyses of model fit using ROC curves and precision-recall curves. The package also contains two functions to create publication-ready tables summarizing model results with an assessment of substantively meaningful effect sizes.

Questions? E-mail me.