Bias-corrected estimators for monotone and concave frontier functions

For the estimation of a monotone and concave support-boundary the data envelopment analysis (DEA) estimator is popular. Recently, under the assumption that the density at boundary is bounded away from zero, Gijbels et al. (J. Amer. Statist. Assoc. 94 (445) 220) derives the limit distribution of the DEA estimator and gives a bias-corrected estimator. In this paper, we generalize the results in Gijbels et al. (1999) by allowing the density at boundary to be infinite, bounded away from zero or zero. (C) 2002 Elsevier B.V. All rights reserved.