scientific article; zbMATH DE number 3976118

From MaRDI portal
Publication:3740829





To investigate two sample problems with a small number of observations on many variables, we examine nonparametric multivariate procedures in which reference distribution of test statistics is calculated on the permutation of observed data. The test statistics are constructed on distance measures defined on a set of all variables. This method has the advantage that the number of variables can be allowed to be much larger than the size of samples. In this paper, we consider four procedures according to the different distance measures. For comparison, we consider approximate F-tests based on the decomposition of variation due to sample, and also the permutation tests. Through applying these methods to an example in the literature and also two examples in the practical fields, we investigate their relative properties. Moreover, through simulation trials, we examine the influence of correlation among the variables and/or heterogeneity of variances on the power of these tests.











This page was built for publication:

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3740829)