|Fallback, Gatekeeping and Tree-Based Strategies for Testing One or More Families of Hypotheses
Brian L. Wiens, PhD
Director, Biometrics and Data Management
Monday, April 24, 2006
E205 Engineering Building
In testing multiple hypotheses in regulated clinical trials, control of the familywise error rate (FWE) is often essential. We consider several methods to control the FWE, many of which are well known when considering a single family of hypotheses. We discuss a newer procedure called the "fallback procedure," a hybrid between the Bonferroni and the fixed sequence procedures. Application of these methods to testing of multiple families of hypotheses is considered using gatekeeping and related approaches. Some extensions of gatekeeping approaches that allow branching ("tree-based" procedures) are introduced.