Estimasi Parameter Regresi Non Parametrik Menggunakan Bootstrap

Abstract

This study aims to estimate non-parametric regression parameters using a spline truncated approach. The approach used is the Ordinary Least Square method and the Bootstrap method. The estimation results are then compared to find out the best non-parametric regression model. The knot points used are 1 knot, 2 knots and 3 knots. Based on the discussion conducted, it is obtained that parameter estimation and non-parametric spline truncated regression model using the Bootstrap method are better than the OLS method. This is because the estimation using the Bootstrap method has a smaller GCV at each number of knot points. The GCV value generated by the bootstrap method is also relatively very small. The difference between the GCV generated by the Bootstrap method and the OLS method is also very large. The best estimate obtained in this study was obtained with 2 knot points using the Bootstrap method. In Bootstrap estimation with 2 knot points and B=50, the GCV value is 4.362. . The best model obtained based on this research is as follows: