Asymptotic Normality of Minimum L sub 1 Norm Estimators in Linear Regression

摘要

It is shown that the minimum L1-norm estimator in linear regression is asymptotically normal with covariance matrix proportional to the inverse of the covariance matrix of the regressors. The method of proof uses asymptotic equicontinuity of Vapnik-Chervonekis graph classes of functions.