Alternative Test Procedures for Testing Homogeneity of the Risk Difference in Multiple Comparative Trials

• Activity Number: 108 - Innovations in Testing and Inference
• Type: Contributed
• Date/Time: Monday, August 8, 2022 : 8:30 AM to 10:20 AM
• Sponsor: Biometrics Section
• Author(s): Krishna K Saha* and Soumik Banerjee
• Companies: Central CT State University and Binghamton University
• Keywords: comparative study; beta-binomial; intra-class correlation; mean-variance; robust variance; size and power

Abstract

Lipsitz et al. (1998) proposed various test procedures for testing the homogeneity of the risk difference in a multicenter randomized clinical trial when the data are sparse. However, in some situations, these test procedures showed serious liberal behaviors. To improve these test procedures, Lui and Kelly (2000) considered three different suggested approaches, but still these test procedures behave very similarly by showing improvement only in power. To overcome these limitations, We propose to improve the existing test procedures using the correct variance estimators by taking within center correlation into account. We also develop some likelihood-based test procedures for testing the homogeneity of the risk difference based on the binomial and the beta-binomial models. Our proposed test procedures are then compared with existing test procedures, by Monte Carlo simulations, in terms of size and power. An illustrative application of the proposed test procedures is presented