THE STUDENTIZED FORM OF THE EXTREME MEAN SQUARE TEST IN THE ANALYSIS OF VARIANCE
成果类型：
Article
署名作者：
NAIR, KR
刊物名称：
BIOMETRIKA
ISSN/ISSBN：
0006-3444
发表日期：
1948
页码：
1631
关键词：
摘要：
In routine analysis of variance, each of k independent estimates of an unknown variance may be tested against another independent estimate so2. The largest and smallest estimates are most likely to be declared significantly different from So2. With k of the order of 20, at least one of the estimates may be expected to be declared significant. A proper test of significance involves the distr. of the largest (or smallest) variance ratio of a set of k. Both distr. functions are obtained for k estimates each having m degrees of freedom and tested against an independent estimate based on [gamma]degrees of freedom. Tables are given of the 5% and 1% points of the distrs. for m = 1 only. These tables are useful for analysis of 2n factorial designs.