Using graphical displays of exact overall significance levels as a guideline for choosing a significance level for pairwise comparisons of a dichotomous variable.

Exact overall significance levels for selected sample sizes and response probabilities are graphically displayed when a dichotomous variable is compared between a placebo and two or more active treatments. A Z-statistic with a pooled variance estimator is used as the test statistic and critical values are based on the (1 - alpha/2)th percentile of the standard normal distribution. These figures are used as a simple guideline for choosing significance levels that achieve a desired overall significance level. If an overall significance level of 0.05 is desired when comparing a placebo to two, three, or four active treatments, the recommended significance levels to be used for each of the two, three, or four pairwise comparisons are 0.027, 0.019, or 0.015. If an overall significance level of 0.10 is desired, the corresponding recommended significance levels are 0.055, 0.039, or 0.031.