Nonparametric Regression

Abstract

Nonparametric regression is a methodology for describing the trend between a response variable and one or more predictors. This approach differs from classical regression models in that it does not rely on strong assumptions regarding the shape of the relationship between the variables. Rather, the data are allowed to speak for themselves in determining the form of the fitted regression functions. Approaches include kernel smoothing and local polynomial regression, spline-based regression models, and regression trees and related techniques. Additive models are typically employed when there are several predictors to mitigate problems arising from the curse of dimensionality.