Generalized biases in nonsymmetric univariate kernels

Abstract

This work extends and generalizes biases in nonsymmetric kernels. The practice of obtaining biases of any nonsymmetric kernel when the order of the smoothing parameter, h, is one is seen not to be sufficient as the error size for this case is large. A new scheme for higher order biases in nonsymmetric univariate kernels is proposed. This scheme enjoys not only the possibility of reducing the size of the global error term (MISE), but also generalizes the bias term for any nonsymmetric kernels.

JONAMP Vol. 11 2007: pp. 497-500