M.S. (Master of Science)
Department of Statistics
Statistics; Public health
The objective of this study is to compare the performance of Bayesian Additive Regression Trees (BART) with Cox Proportional Hazards (CPH) and Random Survival Forests (RSF) models using simulation studies and a real data application on breast cancer survival data as provided by the U.S. SEER database for the year 2005. In the simulation study, we compared the three models across varying sample sizes and censoring rates on the basis of bias and prediction accuracy. Results obtained indicate that the performance of the CPH model depreciates when the PH assumption is violated, however BART continues to perform with almost equal effectiveness. In the real data application, a retrospective analysis was performed in 1500 patients having invasive ductal carcinoma. According to several performance assessment measures, BART and RSF based on log-rank splitting rule fare equivalently and BART marginally outperforms CPH. BART is shown to have similar functioning capacities as RSF, however being in the Bayesian paradigm, BART additionally allows for natural quantification of uncertainty and construction of credible and prediction intervals. The prognostic competence of BART along with the interpretative results obtained using the partial dependence survival functions and variable importance measures can thus be effectually used to solve potential future survival problems.
Saha, Satabdi, "Survival analysis with Bayesian additive regression trees and its application" (2017). Graduate Research Theses & Dissertations. 5158.
viii, 69 pages
Northern Illinois University
Rights Statement 2
NIU theses are protected by copyright. They may be viewed from Huskie Commons for any purpose, but reproduction or distribution in any format is prohibited without the written permission of the authors.
Advisors: Duchwan Ryu.||Committee members: Nader Ebrahimi; Alan M. Polansky.||Includes bibliographical references.||Includes illustrations.