Multiple hypothesis tests in multiple investigations.

Abstract

Inferential statistical methods have traditionally been based on the assumption that one experiment is performed and that interest centres on one or more predetermined hypothesis tests. Exploratory research, on the other hand, often involves multiple hypotheses or repeated investigations under similar or different conditions or both. Several techniques have been proposed to deal with multiple or simultaneous hypothesis testing in single investigations, and procedures to combine observed significance levels for an individual hypothesis test from two or more investigations have been suggested. In this paper we propose a method for identifying important results from multiple statistical tests in multiple investigations. The method is illustrated by using high performance liquid chromatography to identify potential aetiologic contaminants in L-tryptophan samples.