The MLC-CUB model extends the CUB (Combination of a Uniform and shifted Binomial) framework to multivariate rating data.
It assumes that each response results from two latent components:
Feeling – the respondent’s underlying attitude or preference;
Uncertainty – the indecision inherent in the response process.
By combining these components within a model-based clustering approach estimated via the EM algorithm, MLC-CUB identifies groups of respondents sharing similar patterns of feeling and uncertainty.
The figure illustrates this idea, showing clusters characterised by different emotional and decisional profiles.
Ventura M., Jacques J., Zuccolotto P. (2025), Model-based Clustering of Multivariate Rating Data accounting for Feeling and Uncertainty. Journal of Classification
The R functions (MLCCUB_Functions) for fitting and validating the Multivariate Latent Class CUB (MLC-CUB) model are available together with a short user Manual.
These functions implement the EM algorithm for parameter estimation, clustering, and bootstrap-based stability assessment.
The R scripts used to reproduce the application studies presented in the paper will be provided, allowing the replication of the empirical analyses.
Please note that the current version of the code is computationally intensive and may be slow, especially during the bootstrap phase.
An optimized and faster implementation is under development and will be released in a future update.
