Some considerations to carry out a composite indicator for ordinal data

Abstract

Composite indicators (Cis) are very important tools useful in many elds toasses, compare and rank performances, development stage, quality, and manyother dierent targets. CIs are an overall measure of a multi-dimensional, notdirectly observable, concept and are obtained by means of a set of manifestvariables (elementary indicators) which contribute to overall measure of theunobservable concept. In this paper some matters regarding methods tobuild up CIs are discussed assuming elementary indicators are measured onordinal scale and a quantication is necessary to convert observed data intonumerical one. Aggregating function and weighing system are consideredand the use of dissimilarity indexes together with Kendall- -association orheterogeneity measure for weighting is suggested. The suggested procedureis then applied to university students satisfaction data.