Visual Understanding of Higher-Order Kernels

1 December 1994

journal article
research article
Published by Taylor & Francis in Journal of Computational and Graphical Statistics

Vol. 3 (4) , 447-458
https://doi.org/10.1080/10618600.1994.10474657

Abstract

Higher-order kernels have been suggested for use in nonparametric curve estimation. The usual motivation is that they often give faster asymptotic rates of convergence. This article provides a visual derivation of higher-order kernels. This gives new insight into how they work. Furthermore, it is seen that they work well when the curvature of the target curve is roughly constant, and work poorly when there are abrupt changes in curvature on neighborhoods that are the size of the bandwidth.

Keywords

This publication has 7 references indexed in Scilit:

Local Regression: Automatic Kernel Carpentry
Statistical Science, 1993
Exact Mean Integrated Squared Error
The Annals of Statistics, 1992
Choosing a Kernel Regression Estimator
Statistical Science, 1991
Gaussian‐based kernels
The Canadian Journal of Statistics / La Revue Canadienne de Statistique, 1990
Nonparametric Regression Analysis of Longitudinal Data
Published by Springer Nature ,1988
Improvement of Kernel Type Density Estimators
Journal of the American Statistical Association, 1977
Density Estimation for Statistics and Data Analysis
Published by Springer Nature ,1400