Student Seminars

Naveen Naidu Narisetty
PhD student, Department of Statistics, University of Michigan

A Scalable and Consistent Variable Selection Method for High Dimensional Logistic Regression

Within the framework of Bayesian computation, we propose a new variable selection method for logistic regression that adapts to both the sample size n and the number of potential covariates p with desirable features. We use spike and slab priors on the regression coefficients that shrink and diffuse, respectively, as the sample size increases. More importantly, we propose a modified Gibbs sampler whose computational complexity grows only linearly in p, but it retains the property of strong model selection consistency even in the cases of p > n. In contrast with the standard Gibbs sampler, our new algorithm is much more scalable to high dimensional problems, both in memory and in computational efficiency.

Student Seminar Archive

For questions regarding the Statistics Student Seminar or if you are interested in presenting, please contact Joonha Park(joonhap@umich.edu) or Jingshen Wang(jshwang@umich.edu).