Correcting for Edge Effects in Spatial Statistical Analyses

Abstract

Motivation for the research theme discussed in this chapter comes from the myriad of problems arising from dividing geographical landscapes into a finite number of areal units. Some of the more prominent of these problems already have been mentioned in earlier chapters. For example, in Chapter 2 the finite nature of a two-dimensional surface was shown to influence the answer to the question of whether or not data transformations should be employed. In Chapter 3 an empirical relationship was uncovered between the shape of a region and the level of spatial autocorrelation that was measured for its internal geographical distributions. In Chapter 5 the theoretical autocorrelation values for a conditional model, calculated with equation (5.16), were found to diner from those given by the spatial linear operator (I - ϱC)^-1. These differences were found to be most pronounced along the edges. And, in Chapter 6 edge effects were found to be linked to the estimation of values for ϱ that are close to the border of the parameter space. These are but some of the numerous complications introduced into a spatial analysis by edge effects.