Application of geographically weighted regression for assessing spatial non-stationarity

Abstract

Linear regression is a commonly used method of statistical analysis. However, it is not able to capture any spatial variations that may exist in the relationship between explanatory and response variables. We will study geographically weighted regression, which is a local regression method that can account for spatial non-stationarity that may exist. We will describe the model, estimation and hypothesis testing, both in theory and in simulation studies. We will also apply the method to analyze data collected on housing prices in the Boston metropolitan area.