BAYESIAN ANALYSIS
JUNE 25-29, 2018



Welcome to the page for my workshop on Bayesian Analysis at the IPSA-NUS Methods School!

This course provides participants with an applied introduction to Bayesian data analysis and inference. Bayesian methods have rapidly grown in the social sciences in recent years and have become a central tool for a wide variety of analytical methods. Within the Bayesian framework, sampling via Markov chain Monte Carlo (MCMC) methods allows researchers to find solutions to otherwise difficult or intractable statistical problems, and this course offers participants hands-on training in how to use these methods to assess the probability distributions of effect sizes, deal with incomplete data, estimate and incorporate uncertainty in measurement models, use prior information to refine model estimates and predictions, and more.

Covering both Bayesian theory and applications, the course explores the following topics:
  • Why use Bayesian inference?
  • Philosophical and theoretical foundations for Bayesian inference
  • The mechanics of MCMC tools and sampling
  • Building and estimating Bayesian linear and generalized linear models
  • Using MCMC output for post-estimation, incl. marginal effects and predicted probabilities
  • Bayesian approaches to measurement
  • Bayesian tools for model comparison
  • Model presentation and communication
  • Optimal solutions for workflow and reproducibility
Upon completion of this course, participants will be able to:
  • Understand the origins and logic behind Bayesian inference
  • Use Bayesian methods for analyzing continuous and categorical outcomes in a regression framework
  • Use Bayesian methods for measurement models
  • Communicate Bayesian estimation results to practitioners and social science audiences
To allow participants to take full advantage of Bayesian data analysis in their own work, the course also teaches participants how to use the free and open-source software packages R and Stan. Practical examples and applied exercises form an integral part of the course.

Prerequisites

The course presumes a working knowledge of the linear regression model. Familiarity with probability theory would also be helpful, but is not formally required.

Course folder

Click here for the course folder with lecture materials, problem sets, and additional readings. The password will be distributed by the Methods School staff at the beginning of the course.