Abstract:
|
This presentation contributes to the design and analysis in multi-armed randomized clinical trials. We present multi-armed extensions of Efron’s Biased Coin design and the big stick design, which were developed only for two-armed randomized studies, and extensions of randomization-based tests to multi-arm settings, especially the application of the conditional randomization test to the multiple comparisons problem. The view is taken that the linkage of the statistical test to the randomization procedure should be recognized to fulfill immediate utilities (e.g., selection of a randomization procedure in a trial, statistical validity). We provide an efficient algorithm based on Monte Carlo simulation that makes computation of the conditional randomization test feasible, a demonstration of the preservation of type I error rate under heterogeneity and the relation of statistical power with randomization procedure.
|