摘要：It is well known that the ordinary least squares (OLS) regressionestimator is not robust. Many robust regression estimators have been pro-posed and inferential methods based on these estimators have been derived.However, for two independent groups, let j(X) be some conditional mea-sure of location for the jth group, given X, based on some robust regressionestimator. An issue that has not been addressed is computing a 1 con -dence interval for 1(X)2(X) in a manner that allows both within groupand between group hetereoscedasticity. The paper reports the nite sam-ple properties of a simple method for accomplishing this goal. Simulationsindicate that, in terms of controlling the probability of a Type I error, themethod performs very well for a wide range of situations, even with a rela-tively small sample size. In principle, any robust regression estimator can beused. The simulations are focused primarily on the Theil-Sen estimator, butsome results using Yohai's MM-estimator, as well as the Koenker and Bas-sett quantile regression estimator, are noted. Data from the Well Elderly IIstudy, dealing with measures of meaningful activity using the cortisol awak-ening response as a covariate, are used to illustrate that the choice betweenan extant method based on a nonparametric regression estimator, and themethod suggested here, can make a practical di erence.
关键词：ANCOVA; bootstrap methods; Theil-Sen estimator; Well El-;derly II study.
