Abstract
We present a Factor Analysis technique useful for the study of data tables which contain a partition of the individuals set, defined in a natural way. Such a technique allows to exhibit the relations that exist between individuals belonging to the same class, discriminating them from the members of other classes using the relative weights assigned to these classes. By an appropriate selection of the class weights, the new methodology obtains local variables of the Partial Analysis, with the advantage of a representation of the individuals, as active elements, of the local axes.References
Aluja, T., Lebart, L. (1984) “Local and partial principal components analysis and correspondence analysis”, Proceedings COMPSTAT-84, Physica Verlag, Viena: 113-118.
Benali, H.; Escofier, B. (1990) “Analyse factorielle lissée et analyse factorielle des différences locales”, Revue Statistique Appliquée 38(2): 55–76.
Escofier, B. (1984) “Analyse factorielle en référence à un modèle. Application au traitement des tableaux d’échanges”, Revue de Statistique Appliquée 32(4): 25–36.
Escofier, B. (1989) “Multiple correspondence analysis and neighboring relation”, Actes Du Colloque de L’INRIA, Antibes: 55–62.
García-Montoya, M.A. (1998) Tratamiento Factorial de Estructuras definidas mediante Grafos. Aplicación al Estudio de las Tablas Input-Output. Tesis Doctoral, U.P.V./E.H.U., Bilbao.
Holmes, R.B.; Jones, L.K. (1996) “On uniform generation of two-way tables with fixed margins and conditional volume test of Diaconis and Efron”, The Annals of Statistics 24(1): 64–68.