Proposta metodologica per lo studio delle relazioni spaziali tra unità microterritoriali con dati aggregati

Abstract

The cartographic representation of spatial data is a useful tool to highlight of desease/excellence among small areas as regard to an indicator of interest; otherwise possible spatial correlations are not immediatly pointed out in the analysis. 
In this paper we propose a methodology to perform a ranking of small areas (SubT) for level (₁ Y ) and temporary variations  (₂ Y ) of a discretized numeric indicator y, starting from aggregate data (big areas, UT). 
Many methodologies which simultaneously want to estimate the scores of ₁ Y , ₂ Y and the factors which produce the ranking performed are based on principal components, but, in our opinions, other approachs seem more appropriate, for example, predictive methods based on (generalised) regression. Unfortunately this methodology is not applicable in the contest of cartographic representation of a discretized dependent variable and we propose an Ordinal Logit model for 
₁ Y  and ₂ Y : the specified model allows to investigate the role (weight) of predictors in the determination of the ranking performed (on UT) and suggest a way of ranking the SubT.