Permutation methods using median differences for simultaneous pairwise comparisons with a control are investigated. Simulation results suggest that the permutation methods are generally more powerful than the Dunnett procedure when data are from nonnormal distributions. A new procedure is shown...