Bias-robust

The theory of bias-robust regression [ Mar89 ] provides guidelines for the selection of good robust estimators. For calculating beta, either of the following two bias-robust estimators work quite well: (a) an adaptive least trimmed squares estimate, and (b) an MM estimate of [Yoh88], computed as suggested in [Yoh91].

Bias-robust methods for estimating the mean squared error ( MSE ) of linear predictors of finite population quantities, i.e., methods that remain approximately unbiased under failure of assumptions about second order and higher moments, have been developed.