COMING SOON! PQDT Open is getting a new home!

ProQuest Open Access Dissertations & Theses will remain freely available as part of a new and enhanced search experience at

Questions? Please refer to this FAQ.

Dissertation/Thesis Abstract

Contributions to Statistical Testing, Prediction, and Modeling
by Pesko, John Carl, Ph.D., The University of New Mexico, 2017, 136; 10264281
Abstract (Summary)

1. "Parametric Bootstrap (PB) and Objective Bayesian (OB) Testing with Applications to Heteroscedastic ANOVA": For one-way heteroscedastic ANOVA, we show a close relationship between the PB and OB approaches to significance testing, demonstrating the conditions for which the two approaches are equivalent. Using a simulation study, PB and OB performance is compared to a test based on the predictive distribution as well as the unweighted test of Akritas & Papadatos (2004). We extend this work to the RCBD with subsampling model, and prove a repeated sampling property and large sample property for general OB significance testing.

2. "Early Identification of Binswanger's Disease Patients Using Random Forests": We use cross validation to compare several methods for predicting if vascular dementia patients are of the Binswanger type or if they more likely suffer from some other small vessel disease. We investigate which biomarkers are most important for classification, and see that a random forest algorithm accurately identifies Binswanger's patients years before a clinical diagnosis can be ascertained.

3. "High-Throughput Gene Expression Analysis Under the Case-Cohort Study Design": The case-cohort study design blends the efficiency of case control studies with the philosophical soundness of full cohort studies, and presents an efficient way to analyze survival data, particularly for large cohorts with low failure rates. Using a tandem of real data examples and simulation studies, we investigate the performance of the most popular case-cohort analysis approaches in the context of high-dimensional biomarker evaluation.

Indexing (document details)
Advisor: Zhang, Guoyi
Commitee: Christensen, Ronald R., Kang, Huining, Lu, Yan
School: The University of New Mexico
Department: Statistics
School Location: United States -- New Mexico
Source: DAI-B 79/01(E), Dissertation Abstracts International
Subjects: Biostatistics, Statistics
Keywords: Bayesian significance testing, Binswanger's disease, Case-cohort design, High-throughput gene expression, Parametric bootstrap, Random forests
Publication Number: 10264281
ISBN: 978-0-355-15663-8
Copyright © 2021 ProQuest LLC. All rights reserved. Terms and Conditions Privacy Policy Cookie Policy