文章基本信息

标题：Some new tables of the largest root of a matrix in multivariate analysis: a computer approach from 2 to 6 - Section on Statistical Computing - Brief Article
作者：William Chen
期刊名称：Statistics of Income Bulletin
印刷版ISSN：0730-0743
电子版ISSN：1945-2608
出版年度：2002
卷号：Winter 2002
出版社：U.S. Internal Revenue Service

Some new tables of the largest root of a matrix in multivariate analysis: a computer approach from 2 to 6 - Section on Statistical Computing - Brief Article

William Chen

The distribution of the non-null characteristic roots of a matrix derived from sample observations taken from multivariate normal populations is of fundamental importance in multivariate analysis. The Fisher-Girshick-ShuRoy distribution, which has interested statisticians for more than six decades, is revisited in this study. Instead of using K.C.S. Phillai's method by neglecting higher order terms of the c.d.f, of the largest root to approximate the percentile points, we simply keep the whole c.d.f., then apply its natural property to find all the needed percentile points. For the duplicated percentile points, we found our results consistent to the existent tabulation. However, we have greatly extended the tables.