Alternate distance metrics in spatial statistics: radial adjustment of vectors in Euclidean networks

Abstract

The propensity of utilizing Euclidean distance metrics when calculating spatial statistics generally ignores the underlying connectivity between the features under analysis. A procedure is developed to compensate for the distance discrepancies inherent in spatial statistics algorithms by temporarily transforming the model features into an alternate distance metric space that more realistically represents the functional connectivity distance between spatial elements. -- Comparative statistical analysis results between the adjusted and un-adjusted spatial arrangements suggest that statistical measures that are strictly distance based can display dramatic differences in the magnitude of these results. Global autocorrelation measures display much less variation while local autocorrelation measures can result in regions of expanded spatial clustering.