.05: A Case of the Tail Wagging the Distribution

Abstract

After a brief review of the origins of the .05 criterion for judging statistical significance, this article challenges the management discipline's rigid adherence to conventional levels of significancefor differentiating reliable from unreliable results. Arguing instead that there is no right or wrong level of significance, it contends that the selection of a significance level should be treated as one more research parameter. In this respect, the appropriateness of a specific level of significance should be based on such considerations as sample size and research design rather than a priori declarations of .05, .01, or whatever.