tests for spatial autocorrelation are designed to quantify the extent of clustering and are measures of similarity between association in value (covariance, correlation, or difference) and association in space (contiguity). A listing of spatial autocorrelation statistics for increasing orders of contiguity is referred to as a spatial correlogram.
The presence of strong relationships among observations taken from points in space. It results in biased regression coefficients. Special statistical techniques, known as Spatial Statistics and Spatial Econometrics, need to be applied to correct the problems associated with spatial autocorrelation.
The degree to which a set of features tend to be clustered together (positive spatial autocorrelation) or be evenly dispersed (negative spatial autocorrelation) over the earth's surface. This is often measured using either Geary's coefficient or Moran's coefficient. When data are spatially autocorrelated the assumption that they are independently random is invalid, so many statistical techniques are invalidated.