Online Program Home
My Program

Abstract Details

Activity Number: 258
Type: Contributed
Date/Time: Monday, August 1, 2016 : 2:00 PM to 3:50 PM
Sponsor: Section on Bayesian Statistical Science
Abstract #321272
Title: Bayesian Single Index Models
Author(s): Kumaresh Dhara* and Debajyoti Sinha and Debdeep Pati
Companies: Florida State University and Florida State University and Florida State University
Keywords: Bayesian Statistics ; Single Index Model ; Mode Alignment ; Markov Chain Monte Carlo ; Gaussian Process Prior

Single index model is a popular tool in a wide variety of applications ranging from biomedical research to machine learning. From a frequentist viewpoint, there is a substantial amount of literature devoted to proposing strategies for estimating and inferring parameters of the model, with corresponding theoretical justification regarding consistency and asymptotic efficiency. On the other hand, there are currently only a handful articles on Bayesian single index models. In a Bayesian setup, we need a suitable prior process for the unknown nonparametric univariate function and another prior distribution on the sphere for the unknown coefficients. While using these types of priors, the posterior distribution of the coefficients does not have a closed form expression, and existing tools typically results in slow mixing of the Markov chain. In this article, we propose an efficient variant of the Metropolis Hastings algorithm to sample from the full conditional distribution using an Orsntein Uhlenbeck (OH) process for the nonparametric effect and a carefully chosen proposal density based on model alignment. The use of OU process results in fast convergence of the Markov chain.

Authors who are presenting talks have a * after their name.

Back to the full JSM 2016 program

Copyright © American Statistical Association