Premium
Some rank tests for one–sided many–one comparisons
Author(s) -
Rothe G.
Publication year - 1989
Publication title -
statistica neerlandica
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.52
H-Index - 39
eISSN - 1467-9574
pISSN - 0039-0402
DOI - 10.1111/j.1467-9574.1989.tb01251.x
Subject(s) - nonparametric statistics , wilcoxon signed rank test , mathematics , rank (graph theory) , statistics , sample (material) , constraint (computer aided design) , sample size determination , power (physics) , econometrics , combinatorics , mann–whitney u test , geometry , chemistry , chromatography , physics , quantum mechanics
We compare some nonparametric tests for the (/+ 1)–sample problem with additive effects under the constraint that in every sample the treatment effect is not less than that in the first sample, i.e. of some control. The behavior of the Pitman efficiency of the respective tests (essentially tests of a Kruskal–Wallis–, Wilcoxon–, Fligner–Wolfe–, Steel–, and Nemenyi–type) is discussed which turns out to depend on the level and power of the tests as well as on the directions, from which the alternative tends to the hypothesis. It will be shown that none of the tests under consideration is uniformly superior to the others.