Weighted Randomization Test for Multi-Arm Randomized Clinical Trials

• Activity Number: 555
• Type: Invited
• Date/Time: Wednesday, August 4, 2010 : 2:00 PM to 3:50 PM
• Sponsor: General Methodology
• Author(s): Feifang Hu*+
• Companies: University of Virginia
• Keywords: randomization sequence ; computational feasibility

Abstract

Randomization test is a standard non-parametrical method for comparing treatment effect in two-arm randomized clinical trials. However, it is unclear how to conduct randomization tests for multi-arm randomized clinical trials, because in the standard randomization test, treatment groups other than those of interest are also involved in a random re-arrangement process, and it is computationally impossible to generate enough useful randomization sequences. To overcome these difficult, we propose weighted randomization test for multi-arm randomized clinical trials. The weighted randomization test is theoretically well founded and can be easily implemented in practice due to its computational feasibility. Also the weighted randomization test reduces to the standard randomization test for two-arm randomized clinical trials. Some numerical studies are also presented.