Online Program Home
My Program

Abstract Details

Activity Number: 254 - Contributed Poster Presentations: Section on Bayesian Statistical Science
Type: Contributed
Date/Time: Monday, July 29, 2019 : 2:00 PM to 3:50 PM
Sponsor: Section on Bayesian Statistical Science
Abstract #307074
Title: Sparse Priors for Orthogonal Matrices
Author(s): Michael Jauch* and Peter Hoff and David Dunson
Companies: Duke University and Duke University and Duke University
Keywords: Bayesian; Multivariate data ; Orthogonal matrices; Sparse; Stiefel manifold

Statistical models for multivariate data are often naturally parametrized by the set of orthogonal matrices. In such models, we may expect that an orthogonal matrix parameter is nearly sparse, i.e. that its entries are mostly near zero with a small fraction of relatively large values. We show how one can construct continuous sparse prior distributions for an orthogonal matrix parameter via the polar decomposition of an unconstrained random matrix. Our family of prior distributions has appealing theoretical properties and is amenable to posterior simulation in standard software. We illustrate its use in applications to gene expression and audio signal processing data.

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

Back to the full JSM 2019 program