Analysis of Meteorological Data Using Nonparametric Multiple Regression

• Activity Number: 424
• Type: Contributed
• Date/Time: Wednesday, August 5, 2009 : 8:30 AM to 10:20 AM
• Sponsor: Section on Nonparametric Statistics
• Author(s): Ozlem Alpu*+ and Berna Yazici and Betul Kan
• Companies: Anadolu University and Eskisehir Osmangazi University and Anadolu University
• Keywords: Multivariate adaptive regression splines (MARS) ; thin plate regression splines (TPRS) ; cubic regression splines ; nonparametric regression ; semiparametric regression

Abstract

Nonparametric regression analysis relaxes the assumption of linearity, substituting the much weaker assumption of a smooth population regression function. Hence, potentially a more accurate estimate of the regression function is provided. In this study, different parametric, semiparametric and nonparametric regression methods are applied on meteorological data set which takes account the clearness index of Eskisehir, Turkey. In order to determine the relationship between the variables, multivariate adaptive regression splines (MARS), thin plate regression splines (TPRS), cubic regression splines are used. The models are compared due to their adjusted determination coefficients (R^2) and deviance values, Akaike Information Criterion (AIC), Generalized Cross Validation (GCV). The results are also interpreted graphically.