Propriétés et interprétation de la covariance relationnelle en ACP
Résumé
This paper is dedicated to the study of the main properties of the so called 'Relational Principal Components Analysis' (RPCA), that achieves the analysis of a random vector, with respect to the prior knowledge of one binary relationship upon the underlying probabilistic space. We detail the relational covariance and expectation properties that are the grounds of this technique, which whilst not being novel, remains scarcely studied. The paper presents with didactic examples for the properties we previously addressed and throw some light on interpretations in RPCA