Masking unmasked
本文提出一种基于最小中位数平方回归的揭蔽方法，用于识别多重回归中因掩蔽效应而无法被单点删除诊断发现的多个异常值，并辅助后续的多点删除诊断确认。
Diagnostic methods based on the deletion of single observations are well established in multiple regression analysis. Multiple deletion methods are also well developed, but are little applied due to combinatorial problems. But sometimes the pattern of multiple outliers cannot be revealed by single deletion methods. In such cases ‘masking’ is said to occur. The method of unmasking described in the paper uses samples of elemental sets of the observations to fit least median of squares regression to the data. This robust method has high resistance and serves as an exploratory tool for the identification of outliers. The techniques of multiple deletion regression diagnostics are then directly applicable for confirmation of the presence of outliers and influential observations.