摘要：AbstractThis study compares the performance of various spatial prediction models which consider spatial dependence among the real estate data in spatial econometrics, spatial statistics, and semiparametric statistics. Thus far, surprisingly few researches have been conducted from this perspective. This study employs the dataset of apartment rent in the Tokyo 23 wards for empirical comparison. This study in particular focuses on a geoadditive model which considers both spatial dependence and nonlinearity of a hedonic function, and suggests the predictive performance of this method superior to those of the traditional methods like kriging or EM type algorithm.
关键词：Spatial prediction;apartment rent;spatial hedonic model;nonlinearity;geoadditive model ;
