Statistical theory of prophylactic and therapeutic trials. II. Methods of operational advantage.

Abstract

Statistical procedures may subserve either of two ends. In the conduct of government, commerce, and manufacture it may be legitimate to invoke them with no aim other than to prescribe a course of action which limits certain assignable risks. We speak appropriately of any such prescription as conditional. In biological research our primary concern is to establish propositions worthy to take their place in the corpus of scientific knowledge accepted as a basis for subsequent action unrestricted by immediate administrative preoccupations. We speak of any such assertion as unconditional. Much needless confusion concerning the credentials of statistical techniques arises through failure to recognize how far each is meaningful in one or other domain. Since our concern in this context is with the validification of results obtained in the conduct of scientific research, unconditional statisti cal inference alone is relevant to the end in view.