L'aggregazione dei dati territoriali ed i suoi effetti sui metodi statistici

Abstract

Spatial zonal data are constituted by aggregation of the characteristics of individuals. There is a wide range of different ways in which this aggregation can be done. The problem of the arbitrariness and the irregularity of areal units is referred to as the modifiable areal unit problem. In studying the effects of aggregation and scale on the statistical properties of some univariate and multivariate parameters, complex analytical expressions have been recently found for spatial processes related to aggregated areal units in a very simple kind of spatial autoregressive model reasonable for well behaved variables (Arbia 1988). If the data we are studying follow a process whose distribution is far from being normal, a Monte Carlo approach can be very useful in studying, at least empirically, the effects of spatial data configuration on statistical methods. Random aggregations of n zones in m (where m < n), with a contiguity constraint, have been performed on three different spatial systems. Zoning distributions of some ordinary statistics have been evaluated for data generated from four different probability distributions. This approach has proved to be very useful to understand important peculiarities of statistical methods in spatial data analysis.