Activity Number:
|
64
|
Type:
|
Topic Contributed
|
Date/Time:
|
Monday, August 12, 2002 : 8:30 AM to 10:20 AM
|
Sponsor:
|
Biopharmaceutical Section*
|
Abstract - #300135 |
Title:
|
Multiple Decision Making in Adaptive Designs
|
Author(s):
|
Peter Bauer*+
|
Affiliation(s):
|
Institut fur Medizinische Statistik der Universitat Wien
|
Address:
|
Schwarzspanierstr 17, Wien, International, A-1090, Austria
|
Keywords:
|
Adaptive designs ; multiple inference ; many-one comparisons ; non-inferiority ; superiority ; switching goals
|
Abstract:
|
Planning an experiment with multiple inferences is a demanding task. Multiple assumptions have to be made for sample size planning. Different questions may require different sample sizes. Some questions may be skipped during the course of the experiment because of internal or external evidence. Early decisions on single questions may occur. Adaptive designs open a wide field of flexibility not requiring a full a priori definition of the adaptation rules. As an example, multiple level alpha tests in adaptive sequential designs are considered for many one-multiple treatment control comparisons. In a different context, multiple decisions with regard to equivalence and superiority of a test treatment--as compared to a control--are considered by applying continuous families of null hypotheses. Here the sequential decision boundaries may vary between the hypotheses. The experimenter may aim to establish non-inferiority early (with narrow rejection boundaries at the beginning) and decide either to stop or to go on for a smaller non-inferiority margin or even for superiority (with narrow rejection boundaries at the end). Finally, some general remarks on adaptive designs are given.
|