B1056
Title: Quantile-based MANOVA: A new tool for inferring multivariate data in factorial designs
Authors: Marlene Baumeister - TU Dortmund University (Germany)
Markus Pauly - Technical University of Dortmund (Germany)
Marc Ditzhaus - Otto-von-Guericke University Magdeburg (Germany) [presenting]
Abstract: Multivariate analysis-of-variance (MANOVA) is well known and applied in all kinds of areas to examine multivariate endpoints. While classical approaches depend on restrictive assumptions like normality and homogeneity, there is a recent trend toward more general and flexible procedures. We proceed on this path but do not follow the typical mean-focused perspective in statistical inference. In contrast, we consider general quantiles, in particular the median, for a more robust analysis against outliers. The resulting, flexible methodology is shown to be asymptotically valid and consistent. The theoretical results are complemented by an extensive simulation study for small and moderate sample sizes, and by illustrative data analysis.
