Digital divide arises from a combination of observable socio-demographic factors and unobservable features, including individuals’ digital attitudes, cognitive skills, and openness to technological innovation. To capture the latent relationships underlying this phenomenon, we develop a model-based clustering approach to group individuals in Italy on the basis of their digital skills. This is done by using a set of binary indicators on digital abilities collected from the 2021 edition of the “Aspects of Daily Life” survey administered by the Italian National Statistical Office. Methodologically, we extend the Mixture of Experts Latent Trait Analyzers (MELTA) model to handle complex survey designs. MELTA is a model-based clustering approach for multivariate binary data that integrates finite mixtures with multidimensional latent traits. It further allows modeling the impact of covariates on cluster allocation, on the conditional responses’ distribution, on both (as in standard mixtures of experts models), or on neither. To deal with survey data, MELTA is extended by formally incorporating sampling weights into both the estimation algorithm and the uncertainty quantification. The proposed framework enables the identification of population-level digital profiles in Italy, while also assessing the role of socio-demographic characteristics in cluster formation and digital skill usage in the target population.
Analyzing digital divide in Italy via Sample-Weighted Mixture of Experts Latent Trait Analyzers / Dalila Failli, M.F.M.. - In: ADVANCES IN DATA ANALYSIS AND CLASSIFICATION. - ISSN 1862-5355. - ELETTRONICO. - (In corso di stampa), pp. 0-0.
Analyzing digital divide in Italy via Sample-Weighted Mixture of Experts Latent Trait Analyzers
Dalila Failli
;Maria Francesca Marino;
In corso di stampa
Abstract
Digital divide arises from a combination of observable socio-demographic factors and unobservable features, including individuals’ digital attitudes, cognitive skills, and openness to technological innovation. To capture the latent relationships underlying this phenomenon, we develop a model-based clustering approach to group individuals in Italy on the basis of their digital skills. This is done by using a set of binary indicators on digital abilities collected from the 2021 edition of the “Aspects of Daily Life” survey administered by the Italian National Statistical Office. Methodologically, we extend the Mixture of Experts Latent Trait Analyzers (MELTA) model to handle complex survey designs. MELTA is a model-based clustering approach for multivariate binary data that integrates finite mixtures with multidimensional latent traits. It further allows modeling the impact of covariates on cluster allocation, on the conditional responses’ distribution, on both (as in standard mixtures of experts models), or on neither. To deal with survey data, MELTA is extended by formally incorporating sampling weights into both the estimation algorithm and the uncertainty quantification. The proposed framework enables the identification of population-level digital profiles in Italy, while also assessing the role of socio-demographic characteristics in cluster formation and digital skill usage in the target population.| File | Dimensione | Formato | |
|---|---|---|---|
|
melta_ADAC-2.pdf
Accesso chiuso
Descrizione: ADAC_2026
Tipologia:
Versione finale referata (Postprint, Accepted manuscript)
Licenza:
Tutti i diritti riservati
Dimensione
557 kB
Formato
Adobe PDF
|
557 kB | Adobe PDF | Richiedi una copia all'autore |
I documenti in FLORE sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.



