Breadcrumb
Mean Field Variational Bayesian Inference for Marginal Longitudinal Semiparametric Regression
Fri 03 May 2013, 16:00
Marianne Menictas
University of Technology Sydney
Organisers: Nick Whiteley, Feng Yu
ABSTRACT
Recent advances in Bayesian inference methods have led to Bayesian approaches to semiparametric regression problems becoming commonplace, in particular Markov chain Monte Carlo (MCMC) methods which can be costly in computation time. The primary aim of this talk is to describe the benefit of using a fast deterministic approach to MCMC known as mean field variational Bayes (MFVB) for the marginal longitudinal nonparametric regression problem and some of its semiparametric extensions. MFVB produces approximate inference, rather than `exact' inference produced by MCMC, however, the approximations are quite pleasing. Extensions to additive, interaction and varying coefficient models are also considered. The methodology is applied in a simulation study and a real data example from a nutritional epidemiology study.