Robust multivariate estimation based on statistical depth filters

Abstract In the classical contamination models, such as gross-error (Huber and Tukey model or case-wise contamination), observations are considered units to be identified outliers not. This is very useful when number of variables moderately small. Alqallaf et al. (Ann Stat 37(1):311–331, 2009) show limits this approach for a larger introduced independent (cell-wise contamination) where now cells One deal, at same time, with both type filter out contaminated from data set then apply robust procedure able handle missing values. Here, we develop general framework build filters in any dimension based on statistical depth functions. We that previous approaches, e.g., Agostinelli (TEST 24(3):441–461, 2015b) Leung (Comput Data Anal 111:59–76, 2017), special cases. illustrate our method by using half-space depth.