Abstract
The paper considers the usual Rank Analysis of Convariance (RANOCOVA) problem, but under the known permutation symmetry of the primary response variates. The setup crops up in many biometric (e.g. ophthalmological studies) as well as engineering applications where the units on which the responses are measured cannot be taken as independent, but may well be assumed to have exchangeable multivariate distributions for proper natural groupings of the responses. In such cases the conventional nonparametric RANOCOVA methods are eminently applicable, but, the present paper proposes a method by which the efficiency of the inferential processes can be increased substantially. The paper also reports the results of a limited simulation experiment performed to compare the proposed method to the conventional RANOCOVA procedure (cf. Sen, 1984).
AMS 1991 Subject classification : 62G 10.
Get full access to this article
View all access options for this article.