Chen, Ling-Wan
(2018)
Regression Models for Dynamic Treatment Regimens and
Quantile Association of Bivariate Survival Data.
Doctoral Dissertation, University of Pittsburgh.
(Unpublished)
Abstract
In this dissertation we propose two new regression models under different types of survival data, including regression analysis for cumulative incidence functions (CIFs) under two-stage randomization, and quantile association regression for bivariate survival data.
The first topic concerns dynamic treatment regimens (DTRs) which are sets of rules for choosing effective treatments for individual patients based on their characteristics and intermediate responses, and have drawn considerable attention in the field of personalized medicine. Sequential Multiple Assignment Randomized Trial (SMART) design is often used to gather data on different DTRs. In this dissertation, we focus on finding personalized optimal DTRs from a two-stage SMART by regressing covariates on CIFs for competing risks outcomes. To our best knowledge no regression is readily available for analyzing competing risks outcome data from a SMART. Thus, we extend existing CIF regression models to handle covariate effects for DTRs. Asymptotic properties are established for our proposed estimators. We show the improvement provided by our proposed methods through simulation studies, and illustrate its practical utility through an analysis of a two-stage neuroblastoma study, where disease progression is subject to competing-risk censoring by death.
In the second project, we focus on local association in bivariate survival times, which is often of scientific importance. The local association measures capture the dynamic pattern of association over time, and it is desirable to quantify local association for different characteristics of the population. In this work, we adopt a novel quanitle-based local association measure, which is free of marginal distributions, and propose a quanitle association regression model to allow covariate effects on the local association under the copula framework. Estimating equations for the quantile association coefficients are constructed via the relationship between this quanitle-based measure and the copula model. To avoid estimating density functions in variance estimation, we extend the induced smoothing idea to our proposed estimators in obtaining the covariance matrix. The asymptotic properties for the resulting estimators are studied. The proposed estimators and inference procedure are evaluated through simulation, and applied to an age-related macular degeneration (AMD) dataset in studying risk factors on the association between AMD progression in two eyes.
Share
| Citation/Export: |
|
| Social Networking: |
|
Details
| Item Type: |
University of Pittsburgh ETD
|
| Status: |
Unpublished |
| Creators/Authors: |
|
| ETD Committee: |
|
| Date: |
25 July 2018 |
| Date Type: |
Publication |
| Defense Date: |
19 July 2018 |
| Approval Date: |
25 July 2018 |
| Submission Date: |
26 June 2018 |
| Access Restriction: |
No restriction; Release the ETD for access worldwide immediately. |
| Number of Pages: |
104 |
| Institution: |
University of Pittsburgh |
| Schools and Programs: |
Dietrich School of Arts and Sciences > Statistics |
| Degree: |
PhD - Doctor of Philosophy |
| Thesis Type: |
Doctoral Dissertation |
| Refereed: |
Yes |
| Uncontrolled Keywords: |
Bivariate survival data; Competing risks; Conditional association; Copula; Fine and Gray;
Induced Smoothing; Inverse probability weighting; Odds Ratio; Quantiles regression; Scheike model; Sequential Multiple Assignment Randomized Trial. |
| Date Deposited: |
25 Jul 2018 19:06 |
| Last Modified: |
26 Sep 2018 22:28 |
| URI: |
http://d-scholarship.pitt.edu/id/eprint/34331 |
Metrics
Monthly Views for the past 3 years
Plum Analytics
Actions (login required)
 |
View Item |
![[feed]](/images/feed-icon-32x32.png){kind=icon}