摘要

Consider nonparametric problems of estimating an unknown distribution function, F, under the loss L(F, a) = integral \F(t) -a(t)\2(F(t))alpha(1-F(t))beta dF(t), where alpha is-an-element-of [-1,0] and beta is-an-element-of [-1,1]. It is proved that the Empirical Distribution EDF) is admissible (extending a result of Brown, 1988). Among them, an important case is the loss L(F, a) = integral \F(t) - a(t)\2 dF(t).