Comments on: Robust estimation of multivariate location and scatter in the presence of cellwise and casewise contamination

Abstract

The paper by Agostinelli, Leung, Yohai and Zamar (henceforth ALYZ) marks an important step forward in that it proposes an estimator that is specifically designed to tackle the combination of cellwise (ICM) and casewise (THCM) outliers. As it was argued in the fast growing literature on cellwise contamination (e.g. Alqallaf et al. 2009, Van Aelst et al. 2012, Ollerer et al. 2013, Farcomeni 2014), cellwise outliers form a serious real-life problem which needs to be addressed. And whereas cellwise contamination is difficult enough by itself, the realistic situation in which both cellwise and casewise outliers may occur in the data is even harder. The authors make a point that the methods developed so far can handle one type of outliers or the other, but not yet both.