Tests of significance considered as evidence
成果类型：
Article
署名作者：
Berkson, J
刊物名称：
JOURNAL OF THE AMERICAN STATISTICAL ASSOCIATION
ISSN/ISSBN：
0162-1459
DOI：
10.2307/2279000
发表日期：
1942
页码：
325-335
关键词：
摘要：
In the dominant school of statistics, the test of significance not only characterizes the method of statistics but is considered to be the very essence of all exptl. science. The argument as usually presented (if the experience at hand would occur very infrequently in a given hypothesis, the hypothesis is considered disproved) has an apparent plausibility, but reflection has led to the conclusion that it is erroneous. It seems to be basically illogical, and not in accord with the mode of reasoning in rational discourse. If an event has occurred, the definitive question is not, Is this an event which would be rare if the null hypothesis is true?, but, Is there an alternative hypothesis under which the event would be relatively frequent? If there is no plausible alternative, the rarity is irrelevant to a decision. In case the event is of frequent occurrence under the null hypothesis, this is to be taken not merely as lack of evidence against but as prima facie evidence in favor of the null hypothesis. Whether such evidence is of sufficient weight to be convincing depends partly upon the size of the sample; for, though the rejection of the null hypothesis in case P is small is more or less independent of sample size, if actually it is the fact that conclusions will be drawn from P''s which are not small, then only very considerable numbers in the sample are reliable. It would be desirable to investigate the implications of middle P''s.