地理加权回归模型方法与研究新进展

Research Progress on Geographical Weighted Regression Models:A Perspective of Literature Reviews

地理加权回归模型是现代空间计量经济学中局部分析的一种重要方法。本文基于文献述评的方式,从空间权重矩阵的设定、参数估计、多重共线性的识别与消除、参数在时空上一致性的假设检验、模型稳健性分析与预测等方面阐释了地理加权回归模型的基本建模范式;并从解释变量及其参数调整、被解释变量的不同数据特征、空间权重的不同设定、参数估计及模型选择方法调整等四个维度对地理加权回归模型方法的研究新进展进行了系统的梳理和解析;还从时空地理加权回归模型的创新推进、地理加权回归模型族的一般化检验与优选、地理加权回归模型与空间计量全局模型的融合建模、基于MATLAB的代码标准化平台化建设等方面指出了地理加权回归模型方法的发展方向。

Geographically weighted regression(GWR)is an important method of local analysis in modern spatial econometrics.Based on literature reviews,this paper explains the basic modeling paradigm of the GWR model in perspectives of the setting of the spatial weight matrix,parameter estimation,identification and elimination of multicollinearity,hypothesis tests of parameter consistency in time and space,model robustness analysis,and prediction,etc.This paper systematically reviews the research progress of GWR from four dimensions,namely,features of explanatory variables and their parameters,data characteristics of the dependent variable,settings of the spatial weight matrix,methods of parameters estimation,and model selection.This paper also points out the directions in research,such as the innovation research of spatiotemporal GWR,the general hypothesis tests and models selection among all the possible models in the GWR family,the integrated modeling between spatial global and local models,and codes standardization and platform construction based on MATLAB.