Publication Type:

Journal Article

Source:

Communications in Statistics: Simulation and Computation, Taylor and Francis Inc., Volume 43, Number 10, p.2213-2224 (2014)

URL:

http://www.scopus.com/inward/record.url?eid=2-s2.0-84902687765&partnerID=40&md5=3589457bcab822ed544ff7236310731a

Keywords:

Generalized coiflets, Heteroscedasticity, Mathematical models, Nondyadic points, Regression analysis, Sampling, Vanishing moment, Variance estimate

Abstract:

A wavelet approach is presented to estimate the variance function in heteroscedastic nonparametric regression model. The initial variance estimates are obtained as squared weighted sums of neighboring observations. The initial estimator of a smooth variance function is improved by means of wavelet smoothers under the situation that the samples at the dyadic points are not available. Since the traditional wavelet system for the variance function estimation is not appropriate in this situation, we demonstrate that the choice of the wavelet system is significant to have better performance. This is accomplished by choosing a suitable wavelet system known as the generalized coiflets. We conduct extensive simulations to evaluate finite sample performance of our method. We also illustrate our method using a real dataset. Copyright © 2014 Taylor & Francis Group, LLC.

Notes:

cited By (since 1996)0

Cite this Research Publication

T. Palanisamy and Ravichandran, J., “Estimation of variance function in heteroscedastic regression models by generalized coiflets”, Communications in Statistics: Simulation and Computation, vol. 43, pp. 2213-2224, 2014.