A step-down procedure for multiple tests of treatment versus control in each of several groups

Cheung and Holland extended Dunnett's treatment versus control multiple comparison procedure to the case of several groups. Dunnett and Tamhane provided a step-down procedure more powerful than that of Dunnett for situations where the researcher's interest is in testing and not in simultaneous confidence interval estimation. In this paper we show how one can use the approach of Dunnett and Tamhane to devise a testing procedure for the multi-group problem with power greater than that of Cheung and Holland. We illustrate the new procedure with reanalyses of published data from a clinical experiment.