A new study for understanding the scale effect in the regression analysis of spatial data

Authors: Wangshu Mu*, Arizona State University School of Geographical Sciences and Urban Planning, Daoqin Tong, Arizona State University School of Geographical Sciences and Urban Planning
Topics: Quantitative Methods, Spatial Analysis & Modeling, Geographic Theory
Keywords: Scale, linear regression, MAUP, spatial autocorrelation
Session Type: Paper
Day: 4/6/2019
Start / End Time: 1:10 PM / 2:50 PM
Room: Roosevelt 4, Marriott, Exhibition Level
Presentation File: No File Uploaded


Spatial analysts have long been puzzled by issues relating to scale. It is common to generate vastly different coefficients/results even for the same model when analysis is conducted at different scales. As a critical component of the modifiable areal unit problem (MAUP), many researches have studied this phenomenon. However, there is no consensus on this issue as the literature has led to some conflicting/inconclusive results and explanations. This paper provides a new study to understand and quantify the scale effect in the regression analysis involving spatial data. With both theoretical and simulation analysis, our research shows that the spatial autocorrelation of residuals, as a result of the misspecification of the model, is responsible for the scale effect in regression analysis. The results differ from previous studies that attribute the scale effect to the spatial autocorrelation of data. Our findings can be used to explain the conflicting results in the existing research. We also provide discussion on strategies to eliminate/minimize the modifiable areal unit problem in regression analysis for generating scale free results.

Abstract Information

This abstract is already part of a session. View the session here.

To access contact information login