Research on an anisotropic bandwidth geographically weighted regression model
Bandwidth is the key to the estimation of the independent variable coefficients of geographically weighted regression models.The local independent variables with spatial correla-tion show various potential spatial heterogeneity by determining the most suitable bandwidth.Traditional geographically weighted regression models,including GWR,SGWR and MGWR,all adopt isotropic bandwidth for their independent variable coefficients.In this study,the con-straint assumption about isotropic bandwidth is relaxed,an anisotropic bandwidth geographically weighted regression(AGWR)model is proposed,and the coefficient estimation process and the conditions of anisotropic bandwidth discrimination are given.Taking the fitting process of the influence factors of housing price in South Jiangsu and North Jiangsu of China as an example,the AGWR model is compared with other isotropic bandwidth models.The AGWR model shows good fitting ability.In Monte Carlo simulation,the AGWR model still maintains this advantage compared with other isotropic bandwidth models.The proposed AGWR model provides a basis for the quantitative study of anisotropy spatial difference problem to set up a suitable model form.
NETL
NSTL
万方数据
