OPTIMAL RATES FOR FINITE MIXTURE ESTIMATION
Type de document :
Pré-publication ou Document de travail
Titre :
OPTIMAL RATES FOR FINITE MIXTURE ESTIMATION
Auteur(s) :
Heinrich, Philippe [Auteur]
Laboratoire Paul Painlevé - UMR 8524 [LPP]
Kahn, Jonas [Auteur]
Laboratoire Paul Painlevé - UMR 8524 [LPP]
Laboratoire Paul Painlevé - UMR 8524 [LPP]
Kahn, Jonas [Auteur]
Laboratoire Paul Painlevé - UMR 8524 [LPP]
Mot(s)-clé(s) en anglais :
strong identifiability.
rate of convergence
mixture model
mixing distribution
Wasserstein metric
maximum likelihood estimate
convergence of experiments
Local asymptotic normality
pointwise rate
Superefficient estimator
rate of convergence
mixture model
mixing distribution
Wasserstein metric
maximum likelihood estimate
convergence of experiments
Local asymptotic normality
pointwise rate
Superefficient estimator
Discipline(s) HAL :
Mathématiques [math]/Statistiques [math.ST]
Résumé en anglais : [en]
We study the rates of estimation of finite mixing distributions, that is, the parameters of the mixture. We prove that under some regularity and strong identifiability conditions, around a given mixing distribution with ...
Lire la suite >We study the rates of estimation of finite mixing distributions, that is, the parameters of the mixture. We prove that under some regularity and strong identifiability conditions, around a given mixing distribution with $m_0$ components, the optimal local minimax rate of estimation of a mixing distribution with $m$ components is $n^{-1/(4(m-m_0) + 2)}$. This corrects a previous paper by \citet{Chen} in The Annals ofStatistics.By contrast, it turns out that there are estimators with a(non-uniform) pointwise rate of estimation of $n^{-1/2}$ for all mixing distributions with a finite number of components.Lire moins >
Lire la suite >We study the rates of estimation of finite mixing distributions, that is, the parameters of the mixture. We prove that under some regularity and strong identifiability conditions, around a given mixing distribution with $m_0$ components, the optimal local minimax rate of estimation of a mixing distribution with $m$ components is $n^{-1/(4(m-m_0) + 2)}$. This corrects a previous paper by \citet{Chen} in The Annals ofStatistics.By contrast, it turns out that there are estimators with a(non-uniform) pointwise rate of estimation of $n^{-1/2}$ for all mixing distributions with a finite number of components.Lire moins >
Langue :
Anglais
Commentaire :
48 pages, 1 figure, soumis à The Annals of Statistics en séparant l'article principal (30 pages) et les appendices (19 pages). «Minimax rates for finite mixtures», hal-01142343, ou arXiv:1504.03506, est une ancienne version, sans aucun résultat point par point, avec beaucoup moins de bibliographie et d'explications, et une présentation différente
Collections :
Source :
Fichiers
- 1507.04313
- Accès libre
- Accéder au document