Most of the currently used discrete nonparametric priors are, with the exception of the Dirichlet process, inconsistent when used to model directly continuous data. On the other hand, when specified as basic building blocks within hierarchical mixture models, they generally lead to consistent density estimation. In this paper we announce several asymptotic results for a large class of discrete nonparametric priors, namely the class of Gibbs-type priors, which will be extensively presented and proved in De Blasi et al. (2011). Specifically, we provide sufficient conditions for consistency and present two examples within this class which exhibit completely opposite asymptotic posterior behaviours when the ''true'' distribution is continuous.

On consistency of Gibbs-type priors

DE BLASI, Pierpaolo;PRUENSTER, Igor
2012

Abstract

Most of the currently used discrete nonparametric priors are, with the exception of the Dirichlet process, inconsistent when used to model directly continuous data. On the other hand, when specified as basic building blocks within hierarchical mixture models, they generally lead to consistent density estimation. In this paper we announce several asymptotic results for a large class of discrete nonparametric priors, namely the class of Gibbs-type priors, which will be extensively presented and proved in De Blasi et al. (2011). Specifically, we provide sufficient conditions for consistency and present two examples within this class which exhibit completely opposite asymptotic posterior behaviours when the ''true'' distribution is continuous.
58th World Statistics Congress of the International Statistical Institute
Dublin
August 21st-26th, 2011
Bulletin of the International Statistical Institute Proceedings of the 58th World Statistics Congress 2011, Dublin
International Statistical Institute
3151
3157
9789073592339
http://2011.isiproceedings.org/
Bayesian Nonparametrics; Asymptotics; Consistency; Gibbs-type priors
DE BLASI P.; LIJOI A.; PRUENSTER I.
File in questo prodotto:
Non ci sono file associati a questo prodotto.

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/2318/89894
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus ND
  • ???jsp.display-item.citation.isi??? ND
social impact