Seminar Series: Guangyong Zou
Date: March 05
Time: 1:30 p.m. EST
"Design and analysis of randomized controlled trials by winning"
Guangyong Zou, PhD
Professor
Department of Epidemiology and Biostatistics
Western University
Abstract:
Randomized controlled trials are the gold standard in the evaluation of interventions. The pretest-posttest design is ubiquitous in research in many fields. With such a design, researchers measure an outcome variable prior to randomization and again as an outcome assessed at follow up. When the outcome is continuous, analysis of covariance (ANCOVA) is the method of choice to deal with accidental bias and improve efficiency. However, application of ANCOVA is nonsensical when the outcome measures are ordinal or continuous determinations that are not compatible with an interval scale because the data cannot be added or subtracted. In this talk, we discuss nonparametric methods for pretest-posttest design with strictly ordinal measurements. We quantify treatment effects using the win probability, defined as the probability that a random subject in the treatment group has a more favorable outcome than that of a random patient in the control group. We propose methods for confidence interval estimation for the win probability and develop sample size formula for estimating treatment effect with given precision and assurance. Simulation results suggest the methods performed very well. We illustrate the methods and provide SAS code using data from previously published clinical trials.
Short Bio:
Dr. Zou is a professor of Biostatics in the Department of Epidemiology and Biostatistics, a scientist at Robarts Research Institute at Western University, and the Director of Biostatistics at Aliementic Inc (formerly Robarts Clinical Trials Inc). He developed the modified Poisson regression for estimating risk ratios. The paper has been cited over 5000 times since its publication in 2004. He also developed the MOVER (method of variance estimate recovery) that has been widely used in constructing confidence intervals. He is currently conducting research on nonparametric methods for clinical trials.
