It is important in regression analysis to detect the leverage data points that exert an unduly large effect on the least squares estimation. However, in practical situations, the existence of leverage data points is complicated by the presence of multicollinearity. In this paper, a class of estimators generalized shrunken M robust - estimators are defined by mixing the biased estimation technique into the robust M estimator. A numerical example is used to illustrate that, these estimators can not only resist the influence of leverage points but also overcome the effects of multicollinearity.
View on SCiNiTO
Top