Applications of kernel regression estimation:survey

Abstract

The topic of nonparametric kernel regression estimation has been extensively studied in the literature since the formulation of the Nadaraya-Watson kernel estimator in 1964. In its general form, the problem concerns estimation of the regression function g under the model Y = g(x) + e when a paired random sample (X_i, Y_i), i=1,…,n is observed. Smoothing kernel functions have been utilized to design n weighted average estimators of the form which approximates the target function g under various modes of convergence. Due to the simplicity of its form and approximation, kernel regression estimation has been shown to be useful in some statistical problems especially at the stage of exploratory data analysis. The aim of this paper is to survey some recent research articles on the application of this estimation procedure. In addition, we recommend a possible application apropos to some missing data models.

Keywords:

• Kernel regression
• weighted least square
• robust estimates
• regression curves
• censored survival data
• random missing
• heteroscedasticity