Authors: Xiang Ye*, University at Buffalo, Peter A. Rogerson, University at Buffalo
Topics: Spatial Analysis & Modeling, Quantitative Methods
Keywords: modifiable areal unit problem, linear regression, omission error
Session Type: Paper
Start / End Time: 9:35 AM / 10:50 AM
Room: Plaza Court 5, Sheraton, Concourse Level
Presentation File: No File Uploaded
An omission error occurs when some independent variables are missing from a regression model. When individual observations are not available, the modifiable areal unit problem (MAUP) appears with the spatially aggregated data sets. Both omission error and the MAUP can occur simultaneously in regression analyses. In particular, the MAUP causes the bias, due to an omission error, to be less predictable for linear regression models and it distorts bias differently with different spatial configurations. This research analyses the impacts of the MAUP on omission error and shows that coefficient estimates at the aggregate level can be decomposed into three parts: the true coefficient, the individual-level bias, and the aggregate-level bias. The findings fill the gap between empirical studies in geography and theoretical results in econometrics and show that the traditional approaches to the MAUP, such as reporting analyses from multiple spatial configurations, are unhelpful to identify the correct coefficients.