A simple Bayes factor for testing measurement-theoretic axioms
Résumé
A lot of models used to represent preferences or other relations have been characterized in the framework of measurement theory. To assess the empirical validity of one of these models, we can either test the model itself or test the axioms characterizing it. When a model is rejected, the latter approach has the advantage of indicating why it is rejected. The available statistical techniques for testing measurementtheoretic axioms are complex and not always satisficing. In this paper, we present a new statistical technique based on a Bayes factor, very simple and flexible.
Origine : Fichiers produits par l'(les) auteur(s)