Experimentwise error rate

From Wikipedia, the free encyclopedia

In statistics, during multiple comparisons testing, experimentwise error rate is the probability of at least one false rejection of the null hypothesis over an entire experiment. The α (alpha) that is assigned applies to all of the hypothesis tests as a whole, not individually as in the comparisonwise error rate. In two or higher factorial ANOVAs the experimentwise error rate is higher than the familywise error rate (error within a factor or interaction). The experiment is regarded as a better conceptual unit than the family (within a contrast) and most statisticians agree that we should be concerned about the overall experimentwise error rate.