One-sided simultaneous comparisons with more than one control

Simultaneous one-sided confidence intervals are developed for comparing several treatment means to more than one control mean under the assumption of i.i.d. normal errors. These intervals are a generalization of Dunnett's comparisons with a single control. Tables of probability points are given for equal sample size allocation. A-optimal sample size allocation is also investigated. An example from the medical sciences with four treatments and two controls illustrates the technique.