Date of Award

December 2019

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Mathematical Sciences

Committee Member

Christopher McMahan

Committee Member

Andrew Brown

Committee Member

Brook Russell

Committee Member

Xiaoqian Sun

Abstract

This dissertation focuses on developing mixed effects models for large scale and complex data. Our motivating applications involve areas where this data is common, including epidemiological studies, environmental sciences, and genetics. Two key attributes for most of the modeling techniques discussed in this dissertation are that they scale easily to large data and that they achieve full variable selection, which is often a desirable trait in mixed effects models. These attributes are primarily handled in two ways. The first is with carefully constructed latent variables that we introduce to make the posterior distributions more tractable. This allows a Markov chain Monte Carlo (MCMC) sampler to be carried out with Gibbs steps, which results in efficient computation of posterior estimates, especially in large data scenarios. The second is through a decomposition of the covariance matrix associated with the random effects and with the use of spike and slab priors, we can achieve full variable selection in not only the fixed effects, but also the random effects. The finite sample performance of our techniques are assessed through extensive simulations and are used to analyze motivating data sets, which includes data from group testing procedures, human disease surveillance studies, and genetics.

Share

COinS
 
 

To view the content in your browser, please download Adobe Reader or, alternately,
you may Download the file to your hard drive.

NOTE: The latest versions of Adobe Reader do not support viewing PDF files within Firefox on Mac OS and if you are using a modern (Intel) Mac, there is no official plugin for viewing PDF files within the browser window.