The aim of the paper is the development of a method for selecting the best representation of a spatial structure among various alternatives. The paper is divided into four parts. In the first part is discussed the quantitative representation of a spatial structure and the spatial weighted matrices are determined. In the second part Moran's spatial autocorrelation coefficient is presented as well as its expected value and variance. Selection methodology and statistical tests are presented in the third part. The last part is the empirical application of the proposed methodology. Empirical results and the command syntax in SPSS pseudo-language are also included in the fourth part.
