Font Size: 

omposite indicators are becoming one of the most prominent analysis tools, especially in social sciences where the need arises to compare and rank groups of respondents by managing huge and diversified amounts of data. The aggregation of information is a powerful yet incomplete operation since it usually disregards of accounting for \textit{uncertainty}. Uncertainty is here meant as the inherent indeterminacy of any decision process, specifically with reference to the discrete-choice process yielding interviewees to provide an ordinal evaluation out of their latent perception. The class of CUB mixture models for ordinal data is grounded on the probabilistic specification of this component, thus establishing a direct control for heterogeneity. Empirical evidence and methodological studies set this framework as an effective statistical modeling among well-known consolidated theories. In this setting, our contribution proposes a technique to build model-based composite indicators that discloses the role of uncertainty also at an aggregated level. The presentation is lead by applications to real data and comparisons with existing methods.