Title

NUMERICAL STUDY OF COVARIATE ADAPTIVE RANDOMIZED CLINICAL TRIALS WITH QUANTILE REGRESSION

Date of Award

Fall 8-2019

Degree Name

Master of Science (MS)

Advisor(s)

HONGJIAN ZHU, PHD

Second Advisor

HAN CHEN, PHD

Abstract

This project presents a numerical study to examine the validity of quantile regression under stratified permuted block randomization. Statistical simulations of type I error rate and power were performed for cases of dependent and independent stratifying covariates as well as for cases where a third covariate is present but excluded from the analyses altogether. The numerical study examined the effect of excluding stratifying covariates from the analyses and found that this resulted in conservative type I error rates and reduced power. A bootstrapped test was subsequently performed in an attempt to correct for the conservative inference which improved type I error and power altogether, but remained too conservative in scenarios where sample size is low or when regression is performed at extreme deciles.

This document is currently not available here.

Share

COinS