Date/Time
Date(s) - 22/01/2026
1:30 pm - 2:30 pm

Speaker: Dr. Andrew Gelman (Columbia University)

Location: Zoom
Please register online at: https://canssiontario.utoronto.ca/event/cast-seminar-andrew-gelman/

Title: Hierarchical Bayesian modeling as a way of life

Abstract: Hierarchical Bayesian modeling is useful for partial pooling in meta-analysis, small-area estimation, and many other areas in which we want to make dense inferences from sparse data.  Although these methods have been around for several decades, many little-known issues and open questions remain.  We discuss several of these, including meta-analysis with a single study, empirical models for the signal-to-noise ratio, the relation between error variance and group size, anthropic priors, the search for a theoretical justification of weakly informative priors, different measures of influence, why we should care about R-squared, the Vermont/Wyoming problem, different ways to model varying treatment effects, the unification of causal methods in econometrics, and the possibility of a new level of abstraction for expressing multilevel models with interactions.