A unified approach to nonparametric trend tests for dependent and independent samples

被引：5

作者：

Bathke, Arne C. ^{[1
]}

机构：

[1] Univ Kentucky, Dept Stat, Lexington, KY 40506 USA

来源：

METRIKA | 2009年 / 69卷 / 01期

关键词：

Nonparametric model; Rank test; Ordinal data; Independent samples; Dependent samples; Trend test; RANK-SCORE TESTS; FACTORIAL-DESIGNS; HYPOTHESES; ALTERNATIVES; STATISTICS;

D O I：

10.1007/s00184-008-0171-x

中图分类号：

O21 [概率论与数理统计]; C8 [统计学];

学科分类号：

020208 ; 070103 ; 0714 ;

摘要：

We consider testing nonparametric hypotheses against ordered alternatives and propose a new unified approach for dependent and independent samples and factorial designs. The new approach allows for arbitrary underlying distributions, including quantitative and discrete ordinal (ordered categorical), or even binary data. It is compared to procedures available in the literature and applied to different data examples. The new method is not only invariant under monotone transformations of the response, but also under monotone transformations of the weights describing the alternative pattern.

引用

页码：17 / 29

页数：13

共 38 条

[1]

Akritas M.G., 1996, Festschrift on the Occasion of the 65th Birthday of Madam L. Puri, P277

[2] FULLY NONPARAMETRIC HYPOTHESES FOR FACTORIAL-DESIGNS .1. MULTIVARIATE REPEATED-MEASURES DESIGNS [J].