Factor analysis provides a canonical framework for characterizing low-dimensional structures in high-dimensional data. In this talk, the focus is on novel extensions of factor analysis to allow settings in which (1) the same set of variables are measured in different studies and there is interest in inferring shared versus study-specific structure, and (2) multiple different types of high-dimensional data are measured on the same individuals and there is interest in characterizing dependence within and across different data types. Using a Bayesian inference approach, we show promising results in applications, simulations and theory.
Speaker: David Dunson, PhD, Arts and Sciences Distinguished Professor of Statistical Science & Mathematics, Duke University
Zoom Information: Contact Liz Buggs to obtain Zoom Link