Masters candidate in Biostatistics, Xue Gong, will present:
“A Hierarchical Testing Procedure for Settings with Binary Outcomes and Potential Treatment Effect Heterogeneity Across Pre-Defined Strata”
Plan B Adviser: Thomas Murray
Abstract: A simulation study was conducted to compare the power of the Cochran-Mantel-Haenszel test, separate Binomial test, as well as the new developed Hierarchical testing procedure under five plausible true scenarios in a multi-center mGlide clinical trial. Due to the limitations and potential low statistical power of the CMH test and separate Binomial test, we started from the Breslow-Day test to check the assumptions of CMH test and collected a group of p-values to conduct the Hierarchical testing. Two levels of pre-defined threshold (p = 0.05, 0.2) of the hierarchical tests were chosen. To have a better assessment of the treatment effect heterogeneity, we used the power curve to describe the power and Type I error under each specific scenario. Based on the result, we can say that the CMH test procedures displayed a strong power overall for which the Binomial test did not, and the Hierarchical tests displayed strong power under the specific conditions for which the CMH test did not. To perform a high-power analysis, we need to consider the trade-off between the Type I error and the power under different application scenarios.
Refreshments will be served prior to the presentation.