Electronic Theses and Dissertations

Identifier

1003

Author

Wen Feng

Date

2013

Date of Award

12-3-2013

Document Type

Thesis

Degree Name

Master of Science

Major

Mathematical Sciences

Concentration

Statistics

Committee Member

Lih-Yuan Deng

Committee Member

E. Olusegun George

Committee Member

Dale Bowman Armstrong

Abstract

Various simulations have been done to demonstrate how propensity score methods perform in estimating mean difference, proportion, odds ratio and marginal hazard ratio, however, few rigorous simulations have been done to evaluate the performance. We conduct an extensive series of simulations to demonstrate the performance of propensity score methods in estimating conditional hazard ratio. For each propensity method, we characterize the conditions to 7 propensity score scenarios, each contains a different set of variables. For each scenario, we apply four propensity methods to estimate conditional hazard ratio. Meanwhile, we also use multivariate Cox model (conventional Cox model) to estimate conditional hazard ratio, along with a variable indicating treatment assignment. We find all propensity score methods result in biased estimation of conditional hazard ratio. Conventional Cox model yields unbiased estimates of conditional hazard ration only if all outcome related variables are included, otherwise, the bias is similar to those from propensity score method.

Comments

Data is provided by the student.

Library Comment

dissertation or thesis originally submitted to the local University of Memphis Electronic Theses & dissertation (ETD) Repository.

Share

COinS