Unicorns Do Exist: A Tutorial on “Proving” the Null Hypothesis

Introductory statistics classes teach us that we can never prove the null hypothesis; all we can do is reject or fail to reject it. However, there are times when it is necessary to try to prove the nonexistence of a difference between groups. This most often happens within the context of comparing a new treatment against an established one and showing that the new intervention is not inferior to the standard. This article first outlines the logic of “noninferiority” testing by differentiating between the null hypothesis (that which we are trying to nullify) and the “nill” hypothesis (there is no difference), reversing the role of the null and alternate hypotheses, and defining an interval within which groups are said to be equivalent. We then work through an example and show how to calculate sample sizes for noninferiority studies.