Elizabeth Bersson, Ph.D., will present the application of a covariance meta regression extension of Bayesian factor analysis that includes information from features summarizing properties of the different exposures. This technique improves performance and enables shrinkage to more flexible covariance structures, reducing the over-shrinkage problem associated with other Bayesian factor analysis approaches that have the disadvantage of shrinking towards a diagonal covariance, often underestimating important covariation patterns in the data. Results will be illustrated using Toddlers Exposure to SVOCs in Indoor Environments (TESIE) data.

Elizabeth Bersson is a postdoctoral fellow with Tamara Broderick in the Laboratory for Information and Decision Systems at MIT. Previously, she received her PhD in statistical science from Duke University. Her research focuses on developing Bayesian methodology for correlated multivariate data.