Determine the best preliminary method for estimating a non-parametric regression model with a practical application

Abstract

Nonparametric regression technique represents a different way in regression analysis of parametric regression technique, but it does not mean that the use of one method prevent the use of the other. The methods of nonparametric regression can be used to assess the legality of supposed nonparametric model and vice versa, and matching the form of regression curve that we have it from nonparametric regression techniques may suggests the appropriate parametric regression model to be used in future studies.

          The research contains the use of some Smoothing methods, Local Polynomial Kernel (LPK), Spline Regression (SR) and Penalized Spline (PS), to estimate a nonparametric regression model, in order to get that we depends on the results of a simulation study that described in experimental side. The comparison was made between the functions that were used in smoothing by using (MAE) and (MSE) criteria to reach the best estimator that represents the data that generated by simulation. In practical side we used real data from Iraqi Stock Market from Gulf Insurance Co.