稳健方法在线性回归模型中的应用

Robust methods applied in linear regression models

从影响函数和崩溃点角度分析了线性回归模型中最小二乘估计的不稳健性,进而引出M估计这类稳健估计,从理论上分析稳健估计的抗差性,并用R软件对实际数据进行实证研究.结果表明,在处理含有异常点的数据过程中,稳健估计优于最小二乘估计.

This paper considers some estimators in linear regression model, least-squares estimator is Confirmed the lack of robustness by analyzing their influence function and breakdown point, robust estimators such as M-estimator is investigated. In addition, the resistant of the robust estimators are analyzed theoretically and empirical application to the actual data by R software illustrates that robust estimators are significantly superior to least squares estimate when data contain outliers.