Poverty modelling with spline truncated, Fourier series, and mixed estimator geographically weighted nonparametric regression.

Multiple linear regressions using spatial data are developed as Geographically Weighted Regression (GWR). It is used to solve the problem of regression models that do not meet the assumptions of homogeneity caused by the nature of each location. Consequently, the global model is less appropriate for usage. In addition, the regression function for each predictor variable is considered different, so it is possible to use a mixed estimator. The goal of this study is to model poverty data with Geographically Weighted Nonparametric Regression (GWNR). The study focuses on modelling poverty data with three nonparametric regression models on the spline GWNR, Fourier GWNR and Mixed GWNR. The results showed that the mixed GWNR was better than the others based on Mean Square Error (MSE) and R-Square (R2) values. [ABSTRACT FROM AUTHOR]