Rank-Based Meta-Analysis

Date of Award

6-2021

Degree Name

Doctor of Philosophy

Department

Statistics

First Advisor

Dr. Joseph McKean

Second Advisor

Dr. Joshua Naranjo

Third Advisor

Dr. Kevin Lee

Fourth Advisor

Dr. Omer Ozturk

Keywords

Meta-analysis, non-parametric, rank-based, fixed-effect, random-effect, confidence interval

Abstract

Traditional meta-analysis is efficient if the random error structure is normal. Real data, though, are often generated by longer (heavier) tailed random error structure. Such data impair the traditional methods, leading to a loss in efficiency of the traditional meta-analysis. In this research, we developed several robust, rank-based meta-analysis fixed-effect and random-effect procedures in a two independent sample setting, as alternatives to traditional parametric procedures when normality assumption is violated. We show that these rank-based procedures are highly efficient compared to traditional methods for longer or heavier tailed random error structure. These methods are also applied in estimating the platelet count reduction for malaria-infected patients to illustrate proposed approaches using real data.

Access Setting

Dissertation-Abstract Only

Restricted to Campus until

6-15-2031

This document is currently not available here.

Share

COinS