Robust Estimation of Regression Model

Several problems emerging with the studentization of M-estimators of regression model are briefly discussed and reasons for an enlargement of famous Hampel's program are given. Paper then presents a proposal of a scale-equivariant and regression-invariant scale estimator with an acceptable computational complexity. Finally numerical results of estimation of scale for several famous data sets are offered as an empirical evidence of finite-sample behavior of the estimator.