Time: 4:30-5:30 p.m.
Date: Wednesday, October 22, 2025
Speaker: Jeff Miller, Associate Professor in the Department of Biostatistics, Harvard T.H. Chan School of Public Health
Title: Bayesian model criticism using uniform parametrization checks
Abstract: Models are often misspecified in practice, making model criticism a key part of Bayesian analysis. It is important to detect not only when a model is wrong, but which aspects are wrong, and to do so in a computationally convenient and statistically rigorous way. We introduce a novel method for model criticism based on the fact that if the parameters are drawn from the prior, and a dataset is generated according to the assumed likelihood, then a sample from the posterior will be distributed according to the prior. Building upon this idea, we propose to reparameterize all random elements of the likelihood and prior in terms of independent uniform random variables. This makes it possible to aggregate across arbitrary subsets of data points and parameters to test for model departures using classical hypothesis tests for dependence or non-uniformity. We demonstrate empirically how this method of uniform parameterization checks (UPCs) facilitates model criticism, and we develop supporting theoretical results.
Bio: Jeff Miller is an Associate Professor in the Department of Biostatistics at the Harvard T.H. Chan School of Public Health, and co-director of the Biostatistics Masters of Science programs. He received his PhD in Applied Mathematics at Brown University, and was a postdoctoral fellow in the Department of Statistical Science at Duke University. His research focuses on statistical methods for genomics, robustness to model misspecification, efficient algorithms for inference in complex models, frequentist analysis of Bayesian methods, and Bayesian model criticism.