Spatial Bayesian Model Selection with Nonlocal Prior Open Access

Ding, Tianyu (2015)

Permanent URL: https://etd.library.emory.edu/concern/etds/d504rk556?locale=en
Published

Abstract

A common objective of Bayesian spatial variable selection for regression models is to

select important spatially distributed predictors that are strongly associated with the outcome. Many Bayesian variable selection procedures use local priors which do not have posterior variable selection consistency. We propose a novel spatial Bayesian model selection procedure based on nonlocal prior with incorporating spatial depen- dence of predictors. It show been shown that the Bayesian model selection procedure with nonlocal prior enjoys good theoretical properties and achieves better perfor- mance than existing methods. In this thesis, we show that incorporating the spatial dependence between the predictors can improve the nonlocal prior based Bayesian variable selection procedure in terms of both selection accuracy and prediction accu- racy. We demonstrate the advantages of our method via simulation studies and an analysis of the brain imaging data from an Autism study.

Table of Contents

1 Introduction

2 The Model 2.1 Nonlocal Priors with Spatial Dependence 2.2 Normalizing Constant Computation 3 Simulation Studies 4 Real Data Studies 5 Discussion Appendices

About this Master's Thesis

Rights statement
  • Permission granted by the author to include this thesis or dissertation in this repository. All rights reserved by the author. Please contact the author for information regarding the reproduction and use of this thesis or dissertation.
School
Department
Subfield / Discipline
Degree
Submission
Language
  • English
Research field
Keyword
Committee Chair / Thesis Advisor
Committee Members
Partnering Agencies
Last modified

Primary PDF

Supplemental Files