MODEL SPLINE DENGAN ERROR BERKORELASI

Abstract

Spline smoothing is a popular method for estimating the function in nonparametric regression  model. Its performance depends greatly on the choice of smoothing parameters. Many methods of selecting smoothing parameters such as GCV, GML and UBR are developed under the assumption of independent observations. They fail badly when data are correlated. In nonparametric regression, correlated error could be solved by finding weighted estimator and determine the correlation matrix from the error. Estimation of nonparametric function is obtained by minimizing the penalized weighted least-square (PWLS). In this paper, the extension of the GML method to estimate the smoothing parameters and correlation simulataneously is presented. Simulation was conducted to evaluate and to compare the performance of  the original GML and the extended GML method. The extended GML is recommended since it works well in all simulation scheme. This method is also able to illustrate the data concentration data in a continous chemical process