A multiple comparison procedure to control the strong stagewise family error rate in comparing test treatments and a control.

Multiple comparisons are commonly seen in clinical trials and many other fields. An example, which is the focus of this paper, is the comparison of several test treatments (possibly different doses of a compound) with placebo (control). It is well known that steps must be taken to control the type I error rate when multiple testing is performed. We introduce the concept of the strong stagewise family error rate and propose a multiple comparison procedure to control this error rate. The proposed procedure is compared to Dunnett's step-down procedure when there are two test treatment groups and a placebo group.