We address replicability in unsupervised latent variable models, proposing a Multi-Study Factor Mixture Analysis framework. Our proposal extends Multi-Study Factor Analysis by jointly modeling shared and study-specific latent factors, while allowing for latent groups through a mixture distribution in the common factor space. Group-specific mixing proportions vary across studies to assess replication of population heterogeneity. Parameters are estimated via a penalized likelihood using an Expetation Conditional Maximization algorithm. The model is applied to identify latent dimensions in a two-study survey investigating involvment and roles in bullying and latent groups characterized by distinct behavioral patterns.

Uncovering Replicated Latent Structure in Factor Mixture Analysis Through a Multi-study Perspective

Genesin Leonardo;Menardi Giovanna;
2026

Abstract

We address replicability in unsupervised latent variable models, proposing a Multi-Study Factor Mixture Analysis framework. Our proposal extends Multi-Study Factor Analysis by jointly modeling shared and study-specific latent factors, while allowing for latent groups through a mixture distribution in the common factor space. Group-specific mixing proportions vary across studies to assess replication of population heterogeneity. Parameters are estimated via a penalized likelihood using an Expetation Conditional Maximization algorithm. The model is applied to identify latent dimensions in a two-study survey investigating involvment and roles in bullying and latent groups characterized by distinct behavioral patterns.
2026
Statistical Science: From Theory to Applied Research III
SIS-FENStatS 2026
978-3-032-30881-8
File in questo prodotto:
File Dimensione Formato  
Multi_Study_Factor_Mixture_Analysis (3).pdf

embargo fino al 16/07/2027

Tipologia: Accepted (AAM - Author's Accepted Manuscript)
Licenza: Accesso libero
Dimensione 198.25 kB
Formato Adobe PDF
198.25 kB Adobe PDF Visualizza/Apri   Richiedi una copia
Pubblicazioni consigliate

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11577/3615640
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus ND
  • ???jsp.display-item.citation.isi??? ND
  • OpenAlex ND
social impact