What is spatial autocorrelation? Reflections on the past 25 years of spatial statistics

Abstract

To date a clear, simple, and concise definition of spatial autocorrelation has eluded the literature, although an increasing number of publications concerned with this concept have appeared in recent years. The scope of this problem is explored here, in a search for a meaning of spatial autocorrelation that is understandable to the average numerate geographer. The primary finding is that this concept can take on any one of the following nine different meanings: self-correlation attributable to the geographical ordering of data; a descriptor of the nature and degree of certain types of map pattern; an index of the information content latent in geo-referenced data, especially that information overlooked by classical statistical estimators when applied to spatial data series; a diagnostic tool for spatial model misspecifi- cation; a surrogate for unobserved geographic variables; a nuisance in applying conventional statistical methodology to spatial data series; an indicator of the appropriateness of, and possibly an artifact of, areal unit demarcation; a spatial process mechanism; and a spatial spillover effect. Each definition is spelled out, in turn, and either numerically or conceptually illustrated. One purpose of this exercise is to aid in the dissemination of introductory spatial statistical understanding throughout the geography community. The importance of attaining this goal lies in uses and potential abuses of spatial statistics that may accompany the proliferation of GIS throughout the technical community.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 160 citations exist upstream.