Bayesian psychometric modeling and blavaan

Full day short course (9:00am – 5:00pm)

Short course #1

The goal of the session is to provide background on Bayesian psychometric models and estimation methods, and to illustrate the models using real datasets. We will consider factor analysis models, item response models, structural equation models, and two-level variations, with distinctions between these models becoming especially blurry under a Bayesian viewpoint. Specific topics include theoretical background, prior and posterior checking, model estimation, and model extension. The topics will be illustrated via case studies in R, especially focusing on the blavaan package and on Stan. 

Intended Audience

Participants who have some experience with psychometric models and/or with Bayesian statistics are likely to get the most out of the class. We will not assume too much existing knowledge of the models or methods, but we will also not have time to start from scratch. There will be R code and blavaan code that participants are free to run on their laptops during the workshop, but this software is not absolutely required to benefit from the course.

Summary

In the time frame of this session, we will attempt to provide participants with resources to do Bayesian psychometric modeling of their own datasets. This includes background about the models themselves, background about Bayesian methods, case studies involving education datasets, and estimation methods. The session will revolve around the principle that psychometric models are best understood as models of raw data, as opposed to models of covariance matrices. This viewpoint is especially advantageous from a Bayesian point of view because it allows for model checks and summaries that are more intuitive than other popular methods. We will spend a good deal of time on case studies because they are helpful for illustrating complicated ideas. The case studies will include PISA data that are easily accessible to attendees. We will also see code examples of MCMC algorithms.

About the instructor

Ed Merkle (University of Missouri)

 Ed Merkle, PhD