A Note on “A Comparison of Two Methods for Adaptive Multi‐Arm Two‐Stage Design”

Mehta and Kappler compare two approaches to adaptive multi-arm multi-stage testing: Stagewise MAMS based on the inverse normal combination test, and cumulative MAMS based on the conditional error rate approach. They show the difference in power between these approaches and conjecture as to its origin. While the results they report are well known, we would like to add a more differentiated insight into the reasons for the power difference.