Regression analysis is a widely-used technique for measuring relationships between two variables or among a number of variables. This collaborative proposal will enable a spatial econometrician (Anselin) and a spatial statistician (Griffith) to address several methodological issues that arise when regression analysis is used to analyze spatially-distributed phenomena. Some of the assumptions needed to validate regression analysis are not met when the technique is applied to spatial data. For example, adjacent spatial observations are commonly not independent of each other. How such violations of validating assumptions affect the results of spatial regression are largely unknown. The applicants will devise measures and tests that will enable scholars who wish to use regression analysis for spatial data to determine how spatial phenomena affect regression analysis and its results, and they will formulate procedures that can be used to test and validate the results obtained from regression when spatial data are involved. Anselin and Griffith will disseminate the results of their research by distributing software incorporating the methods they develop and by conducting workshops to explain their results to interested scientists.
