On ignoring the heterogeneity in spatial autocorrelation: consequences and solutions
Zehua Zhang Ziqi Li Yongze Song a School of Design and the Built Environment,Curtin University,Bentley,Australiab Department of Geography,Florida State University,Tallahassee,Florida,USAZehua Zhang obtained his PhD degree with Chancellor's Commendation from Curtin University in 2023 and was awarded 2023 Humanities Greg Crombie Postgraduate Work of the Year Award. He is currently a researcher of Geospatial Intelligence Lab and working as a Sessional Academic at Curtin University. His research interests include spatial statistics,spatial data science,GIS and transport and population geography. Email: qi Li is an Assistant Professor in quantitative geography at Florida State University (FSU). His research focuses on the methodological development of spatially explicit and interpretable statistical/machine learning models to investigate human behavior across space and place. He is one of the primary developers of Multi-scale Geographically Weighted Regression (MGWR) and Python Spatial Analysis Library (PySAL). Email: Ziqi.Li@fsu.eduYongze Song is a Senior Lecturer at Curtin University. His research interests include spatial statistics,geospatial intelligence,sustainable infrastructure and sustainable development. He servers as an Associate Editor for the journals GIScience & Remote Sensing,and the International Journal of Applied Earth Observation and Geoinformation. Email: Yongze.Song@curtin.edu.au
DOI: https://doi.org/10.1080/13658816.2024.2391981
2024-08-22
International Journal of Geographical Information Science
Abstract:Spatial autoregressive (SAR) models are often used to explicitly account for the spatial dependence underlying geographic phenomena. However, traditional SAR models are specified using a single SAR coefficient, assuming constant spatial dependence over space. This assumption oversimplifies the situation where the true spatial autoregressive process varies in strength; the consequences of ignoring heterogeneous autocorrelation remain to be discussed. This study proposes a heterogeneous spatial autocorrelation model by extending the spatial lag model (SLM). The new model includes change point detection for identifying patterns of spatially varying autocorrelation strengths, a SAR coefficient matrix for representing heterogeneous spatial autocorrelation, and maximum likelihood estimation for determining multiple SAR coefficients. Monte Carlo simulations demonstrate that the proposed method is effective in modeling SAR processes with heterogeneous autocorrelation patterns, while traditional SLM inflates uncertainties in the regression coefficients when a heterogeneous autocorrelation structure is not accounted for. We further applied the new method to an empirical analysis of traffic crashes in the Greater Perth Area, Australia. The heterogeneous spatial autocorrelation model reduces model RMSE by 42% (compared with traditional SLM). Results from both simulation and empirical studies indicate that spatially varying autocorrelation strengths should be considered for SAR processes and relevant applications.
geography, physical,computer science, information systems,information science & library science