线性回归中多个异常点的诊断

Diagnosis of Multiple Outliers in Linear Regression Model

在线性回归模型的诊断分析过程中，针对多个异常点存在时可能发生的掩盖现象，本文简单地分析了应用经典诊断量时存在的问题，介绍了各种顺序剔除方法及稳健回归方法的一般情况，并对各种方法的效能进行了简要的讨论。

 During the diagnostic analysis of linear model,there always exists socalled “masking effect” when multiple outliers occur. In this thesis we analyze the shortcoming of traditional diagnostics in general;and mention some kinds of sequential deletion methods as well as robust regression methods;furthermore we have a discussion on the power of all kinds of methods metioned. 