Reducing sample sizes when comparing experimental and control groups.

Abstract

In comparing rates for two groups (one control and the other experimental), one may use an historical control to reduce the sample size needed for a test of significance to one-fourth that required if the control is not historical. If there is some doubt as to whether one's control is comparable to the historical control, a test showing the equality of the historical and experimental controls can lead to a halving of the sample required for significance testing. This same argument also applies to the comparison of two means.