Regression analysis is widely used to determine the pattern of the relationship between predictor variables and response variables. Nonparametric regression is a regression analysis in which the regression function is unknown and the response variable is correlated. Splines are polynomials that are segmented and flexible so they can adapt effectively to the local properties of data. The shape of the spline estimation is influenced by the lambda parameter when determining the location of the knots point. This study examines how to develop a spline estimation form in nonparametric regression cases by selecting the optimal node using the methods of Generalized Cross Validation (GCV) and Unisex Risk (UBR. From the data used, the optimal knot point is 0.018556 and 0.017783 for the number of splines 5 and 20, respectively. The results of this study indicate that UBR tends to be smaller than the GCV value.
Estimated Spline in Nonparametric Regression with a Generalized Cross Validation and Unbiased Risk Approach
4 Views
20 Downloads