B0335
Title: Geometry and computation of halfspace depth for scatter matrices
Authors: Stanislav Nagy - Charles University (Czech Republic) [presenting]
Abstract: The scatter halfspace depth (sHD) is an extension of the location halfspace (also called Tukey) depth applicable in the nonparametric analysis of scatter of multivariate distributions. Using sHD, it is possible to define minimax optimal robust scatter matrix estimators for multivariate data. Just as the location halfspace depth, also sHD admits natural geometric interpretations. These insights to develop an exact algorithm are used for the exact computation of sHD in any dimension d.
