Variational Bayesian Inference for regression with missing data

Abstract: 

Bayesian hierarchical models are attractive structures for conducting regression analyses when the data are subject to missingness. However, the requisite probability calculus is challenging and Monte Carlo methods typically are employed. We develop an alternative approach based on deterministic variational Bayes approximations. Both parametric and nonparametric regression are treated. We demonstrate that the variational Bayes can achieve good accuracy, but with considerably less computational overhead. The main ramification is fast approximate Bayesian inference in parametric and nonparametric regression models with missing data. This talk represents joint work with Christel Faes(University of Hasselt, Belgium) and John Ormerod (University of Sydney).

About the speaker: Matt Wand is a Research Professor of Statistics in the School of Mathematics and Applied Statistics at University of Wollongong. In addition to his outstanding contributions to methodology for non-parametric regression and density estimation, he has lately worked on variational approximations in complex statistical models. Besides, Matt is a former member of our Department.