Some asymptotic theory for variance function smoothing

Abstract

A practical method is discussed for determining the amount of smoothing when using the kernel method in nonparametric estimation of smooth curves. Simple selection of the smoothing parameter is suggested. Both homoscedastic and heteroscedastic regression models are considered.

Keywords: Asymptotic, Smoothing, Kernel, Bandwidth, Bias, Variance, Mean squared error, Homoscedastic, Heteroscedastic.

> East African Journal of Statistics Vol. 1 (1) 2005: pp. 9-22