Minimum power divergence estimators, maximum likelihood and exponential families
Résumé
This note introduces the dual representation of the divergence between two distributions in a parametric model. Resulting estimators for the divergence as for the parameter are derived. These estimators do not make use of any grouping nor smoothing. It is proved that all power-type divergences with non negative index induce the same estimator of the parameter on any exponential family, which is the MLE.
Domaines
Méthodologie [stat.ME]| Origine | Fichiers produits par l'(les) auteur(s) |
|---|