Abstract

In survey analysis, the latent perception about investigated items is often measured as an ordinal evaluation: this mechanism produces unobserved fuzziness that may blur the final response. According to CUB model paradigm, the data generating process is disentangled as a blend of real beliefs and indecision with the specification of a parsimonious mixture of feeling and uncertainty. The probabilistic structure of the uncertainty component and the flexibility  of interpretations are the major novelties of the approach: also subjects' covariates can be tested in the model to identify significant response drivers and identify outcome patterns.

The benchmark CUB model can be suitably adapted to encompass typical issues in ordinal data analysis, such as the presence of inflated categories, response styles and overdispersion. The modelling of uncertainty has proven successful in several applications stemming from diverse fields of research: psychology, social and political sciences, marketing, etc. The talk will present this perspective by integrating methodology with empirical evidence. On-going research developments and open issues will be outlined