A Simulation Study of Interval Estimation in Nonparametric Regression Using the Truncated Spline Estimator

Abstract

This study examines interval estimation in truncated spline nonparametric regression using simulated data. The study aims to determine the impact of sample size, variance, and knot points on the performance of the truncated spline estimator. The results show that as the sample size increases, both the Generalized Maximum Likelihood (GML) and Mean Square Error (MSE) values decrease, while the coefficient of determination increases. This study also reveals that increasing the variance leads to higher GML and MSE values, as well as a lower coefficient of determination. Furthermore, the truncated spline nonparametric regression model achieves optimal performance with three knot points. The results showed that the more knot points, the GML and MSE values will decrease, while the coefficient of determination increases. The results of this study show that the determination of sample size, variance, and knot points significantly affects the accuracy and efficiency of the truncated spline nonparametric regression model, allowing it to serve as a reference for applying truncated spline nonparametric regression more effectively to produce a more optimal model that aligns with the characteristics of the data.