A semiparametric additive regression model: Investigation of house price in Eskisehir (Turkey)

Abstract

In this paper, different regression models are obtained to examine relation between house price and house features in Centrum of Eskisehir Province (Turkey). The statistical analyses of this paper indicate that some of explanatory variables affect the response variable parametrically and some of them nonparametrically. Therefore, obtained suitable model has both parametric and nonparametric variables and the model is semiparametric additive regression model. Smoothing spline is used for estimating the model, and the predictors for the smoothing spline are obtained backfitting algorithm. In addition to semiparametric additive model, linear model and semiparametric models are constructed, and these models are compared according to deviance and R2 criteria. Then, it has concluded that the semiparametric additive regression model has given better results than the other models. Smoothing spline is used for estimating the model and the predictors for the smoothing spline are obtained by using backfitting algorithm.