ABSTRACT
Significance tests are designed to keep one from making statements about percentage differences when there is too little evidence to justify the statement. In order to do this, in order to avoid making unjustified claims about the magnitude and importance of an observed difference, one sets a rather high level of significance. By insisting on a high level of significance one increases the possibility of rejecting as insignificant differences which actually do exist. To begin with, traditional tests of significance have been developed to study the probable correctness or incorrectness of single, isolated statements, for example, that Drug A is more likely than Drug B to cause nausea. But it is an essential feature of the exploratory studies in medical education, and of many other empirical investigations in the social sciences, that there is no single hypothesis which can be viewed independently of other hypotheses.
