Surmonter un effet Guttman et classer des données d’enquête. Une alliance des analyses factorielles et des classes latentes

Abstract

Overcoming a Guttman effect and classifying survey data: Combining factorial and latent class approaches. Drawing on the analysis of process preferences (preferences regarding decision-making procedures) in Belgium, France, and Italy, based on a large-scale survey, this article demonstrates how we addressed several methodological challenges to produce an interpretable factorial structure and a robust classification across the three national contexts. The contribution of this article is twofold: first, it introduces and documents two underutilized methods–doubled correspondence analysis and multi-group latent class analysis; second, it assesses the respective strengths and limitations of factorial methods and classification techniques, particularly in terms of identifying inter-variable relationships and establishing meaningful typologies. More broadly, the article highlights the importance of methodological trade-offs in the use of factorial and clustering techniques, while emphasizing the interpretive richness that arises from combining methods rooted in different traditions of data analysis.