Distance-based LISA maps for multivariate lattice data

Abstract

In the context of areal data (a particular case of data with spatial dependence) we propose an algorithm to define spatial clusters. Our proposal is based on distance between the characteristics observed in different areas (individual). Thus it is able to be applied to any kind of observable characteristic on condition that an inter-individual distance can be defined. This way we provide a generalization of the well-known LISA maps that have been widely used for univariate data. We apply our proposals to the results of 2004 Spanish General Elections recorded at 248 neighborhoods in Barcelona.