Two-Stage Selection Procedures Based on Tests.
PURDUE UNIV LAFAYETTE IN DEPT OF STATISTICS
Pagination or Media Count:
Suppose that k new treatments have been developed with the purpose of replacing the standard treatment with the best new one, provided that it is actually an improvement on the standard treatment. In a parametric approach, mainly under the assumption of monotones likelihood ratios, procedures are considered which, at a first stage, screen out inferior treatments through statistical tests at a common level of significance. If none exactly one is not eliminated, none this one will be used as a replacement. Otherwise, if more than one treatment overcomes this screening process, that one of the non-eliminated treatments will be chosen as the replacement which is judged to be the best, after additional data have been observed from the selected treatments. Topics of this paper are the questions of how to choose the terminal decision at the second stage and the tests at the first stage, respectively, and how to implement the appropriate procedures at certain pre-specified performance criteria. Author
- Statistics and Probability