Covariates adjustment for nonparametric tests for two sample comparison
Abstract
Two sample nonparametic tests, e.g. Wilcoxon rank sum test has been widely used in clinical trials. However, few studies evaluated the covariate adjustment to these tests. In my dissertation, I evaluated the empirical power and the type I error rates for the test with covariate adjusting approaches when there is covariate imbalance. I also evaluated the mean square error and 95% coverage probability for the point estimation. In the simulation study, we identified ANCOVA approach is not valid when there is severe covariate imbalance and quantile-stratification is not valid when covariate is extremely imbalanced. We applied the proposed approaches to three Phase II clinical trials, TIME, LateTIME, and FOCUS.
Subject Area
Biostatistics
Recommended Citation
Ye, Jiabu, "Covariates adjustment for nonparametric tests for two sample comparison" (2016). Texas Medical Center Dissertations (via ProQuest). AAI10131752.
https://digitalcommons.library.tmc.edu/dissertations/AAI10131752