Stat Colloquium: Dr. Mahlet Tadesse

Title: Bayesian graph-structured variable selection

Abstract: A graph structure is commonly used to characterize the dependence between variables, which may be induced by time, space, biological networks or other factors. Incorporating this dependence structure into the variable selection process can increase the power to detect subtle effects without increasing the probability of false discoveries and can improve the predictive performance. In this talk, I will present methods we have proposed to accomplish this in the context of spike-and-slab priors as well as global-local shrinkage priors. For the former, we specify a binary Markov random field prior that leverages evidence from correlated outcomes to identify outcome-specific covariates. For the latter, we combine a Gaussian Markov random field prior with a horseshoe prior to perform selection on graph-structured variables. The methods will be illustrated with applications in epigenomic, genomic and transcriptomic studies.