Abstract
Autocorrelation provides a means of characterising spatial structure, yet different kinds of structure are revealed in different spaces. The choice of weights when calculating an autocorrelation statistic is discussed. An examination of autocorrelation in transformed spaces is suggested and illustrated empirically by use of correlograms for German population data and autocorrelation statistics for Swedish population data. Transformations are accomplished by use of multidimensional scaling.

This publication has 20 references indexed in Scilit: