A K -sample procedure for comparatively assessing multivariate association

Presents procedures for extending the technique of A. R. Hakstian et al to the context of 2 or more independent samples, which are compared in terms of overall multivariate association. Following a discussion of the notion of multivariate association and a normalizing transformation for the statistic under consideration, formulas are presented for an overall K -sample significance test and subsequent multiple comparisons. On the basis of empirical investigations using Monte Carlo methods, the proposed test statistic is shown, under the null hypothesis, to follow closely the chi-square distribution, thus permitting adequate Type I error control. An example is presented which illustrates the application of the procedures.