Skip to Main Content
Frequently Asked Questions
Submit an ETD
Global Search Box
Need Help?
Keyword Search
Participating Institutions
Advanced Search
School Logo
Files
File List
22307.pdf (865.79 KB)
ETD Abstract Container
Abstract Header
Bayesian Hierarchical Models for Partially Observed Data
Author Info
Jaberansari, Negar
Permalink:
http://rave.ohiolink.edu/etdc/view?acc_num=ucin1479818516727153
Abstract Details
Year and Degree
2016, PhD, University of Cincinnati, Arts and Sciences: Mathematical Sciences.
Abstract
This thesis considers two types of clustered partially observed data and addresses the challenges via Bayesian hierarchical models. Multivariate current status data is a common type of biomedical data, whose analysis is burdened by correlated partially observed data. Here all event times are censored such that at the time of examination, the only available information is whether or not the event happened. To compound this difficulty, an unobserved heterogeneity caused by clusters of units or individuals is also probable. To address these issues, we propose a Bayesian multivariate Gamma-frailty Cox model with two cumulative baseline hazard functions and the corresponding MCMC algorithms. Partially observed data is also an unavoidable obstacle when dealing with observational data analysis for evaluating treatment effect. Often, a set of variables is missing in the majority of patients and observed in only a small subset. The absence of these important variables or confounders poses a threat to the validity of causal inference. In addition, the patients could also be clustered within different medical centers. To deal with these difficulties, we propose a Bayesian hierarchical approach to model the Bayesian propensity score. The method utilizes information from two different data sources: external validation data and main study data, where external validation data provides additional information on unmeasured confounders. We also considered a Bayesian hierarchical approach to missing data imputation. To demonstrate the efficacy of the proposed methods, two case studies are performed. A bivariate current status cataract dataset is analyzed to investigate the effect of various risk factors on the occurrence of cataracts. We also conduct a comparative clinical effectiveness study of treating children with newly onset juvenile idiopathic arthritis. Where necessary, the methods are compared with the non-hierarchical approach, both in simulation and case studies.
Committee
Seongho Song, Ph.D. (Committee Chair)
Bin Huang (Committee Member)
Emily Kang, Ph.D. (Committee Member)
Siva Sivaganesan, Ph.D. (Committee Member)
Xia Wang, Ph.D. (Committee Member)
Pages
104 p.
Subject Headings
Statistics
Keywords
Clustered Partially Observed Data
;
Survival Analysis
;
Causal Effect
;
Bayesian Hierarchical Models
Recommended Citations
Refworks
EndNote
RIS
Mendeley
Citations
Jaberansari, N. (2016).
Bayesian Hierarchical Models for Partially Observed Data
[Doctoral dissertation, University of Cincinnati]. OhioLINK Electronic Theses and Dissertations Center. http://rave.ohiolink.edu/etdc/view?acc_num=ucin1479818516727153
APA Style (7th edition)
Jaberansari, Negar.
Bayesian Hierarchical Models for Partially Observed Data.
2016. University of Cincinnati, Doctoral dissertation.
OhioLINK Electronic Theses and Dissertations Center
, http://rave.ohiolink.edu/etdc/view?acc_num=ucin1479818516727153.
MLA Style (8th edition)
Jaberansari, Negar. "Bayesian Hierarchical Models for Partially Observed Data." Doctoral dissertation, University of Cincinnati, 2016. http://rave.ohiolink.edu/etdc/view?acc_num=ucin1479818516727153
Chicago Manual of Style (17th edition)
Abstract Footer
Document number:
ucin1479818516727153
Download Count:
279
Copyright Info
© 2016, all rights reserved.
This open access ETD is published by University of Cincinnati and OhioLINK.