COMBINATION OF MULTIPLE SCORES FOR NONPARAMETRIC TESTING IN MULTIVARIATE LINEAR-REGRESSION SET-UP

For the problem of testing absence of regression under the p-variate nonparametric linear regression set-up involving m predictors, standard rank test criteria are in the form of a quadratic form in mp linear rank statistics Different standard tests correspond to different choices of one system of scores for each variable. In this paper we propose two test criteria which are based on simultaneous choice of more than one system of scores for each variable. The criteria are obtained by applying the union-intersection technique in two different ways. It turns out that for either criterion the use of several systems of scores for each variable results in an improvement in the asymptotic power. © 1990, Taylor & Francis Group, LLC. All rights reserved.