Bayesian methods for simultaneous clustering and variable selection

The project aims to extend Bayesian clustering, mixture or factor analysis models to incorporate automatic feature selection. This is a challenging problem, to find latent structure in high-dimensional data whilst simultaneously discovering the variables which best predict this structure. The research is motivated by applications in chronic disease epidemiology, integrating multiple high-dimensional data sets simultaneously, enabling researchers to model mediation effects of intermediate exposures and biological markers. Areas of application may be in epidemiology, health economics or bioinformatics.