Nonparametric and robust ways to make sense of multivariate data samples
Arne Bathke, University of Salzburg

Abstract: Different inferential methods for multivariate data samples are presented. Application is demonstrated using the R package npmv. Unlike in classical MANOVA, multivariate normality is not required for the data. Different response variables may even be measured on different scales (binary, ordinal, quantitative). In addition to permutation tests, F approximations, and bootstrap solutions to test the global hypothesis, we present a multiple testing procedure which identifies significant subsets of response variables and factor levels, while controlling the familywise error rate.