Exploring and characterising irregular spatial clusters using eigenvector filtering

The statistical study of spatial clusters is an important part of the exploratory data analysis toolbox. Spatial autocorrelation and hotspot statistics are now routinely used to better understand the arrangement of variables on maps. Spatial clusters, however, are often not internally homogeneous, but may exhibit interesting spatial heterogeneities. In this contribution, approaches to explore irregular spatial clusters are applied to a recent mapped index of food deserts. Both approaches are based on hotspot and heteroscedasticity measures, but one of the methods additionally uses eigenvector filtering. The results show that the latter contributes to the disclosure and understanding of spatial cluster irregularities.