Geographically Weighted Regression: A Method for Exploring Spatial Nonstationarity

摘要: Spatial nonstationarity is a condition in which simple “global” model cannot explain the relationships between some sets of variables. The nature must alter over space to reflect structure within data. In this paper, technique developed, termed geographically weighted regression, attempts capture variation by calibrating multiple regression allows different exist at points space. This loosely based on kernel regression. method itself introduced and related issues such as choice spatial weighting function are discussed. Following this, series statistical tests considered can be described generally for nonstationarity. Using Monte Carlo methods, techniques proposed investigating null hypothesis that data may global rather than non-stationary one also testing whether individual coefficients stable geographic These demonstrated set from 1991 U.K. census relating car ownership rates social class male unemployment. paper concludes discussing ways extended.