Estimation of clearness index model via CRS, TPRS and mars

Abstract

Nonparametric approach is more flexible than parametric approach in assuming that f belongs to a smooth family of functions. Hence, a nonparametric approach does not require an assumption of linearity. Based on our motivating applications, mainly the approach to nonparametric regression is used on clearness index of Eskisehir, Turkey. In this study, cubic regression splines (CRS), thin plate regression splines (TPRS), multivariate adaptive regression splines (MARS) are defined to explore the shape of the functional relationship of the data by constructing numerous models for each. Thin plate regression spline gives the best results for the model in which monthly average daily extraterrestrial radiation on horizontal surface taken as a parametric component and monthly average soil temperature, monthly average sunshine hours are taken as nonparametric components for this data set