August 2021 Spike and slab Pólya tree posterior densities: Adaptive inference
Ismaël Castillo, Romain Mismer
Author Affiliations +
Ann. Inst. H. Poincaré Probab. Statist. 57(3): 1521-1548 (August 2021). DOI: 10.1214/20-AIHP1132

Abstract

In the density estimation model, the question of adaptive inference using Pólya tree–type prior distributions is considered. A class of prior densities having a tree structure, called spike-and-slab Pólya trees, is introduced. For this class, two types of results are obtained: first, the Bayesian posterior distribution is shown to converge at the minimax rate for the supremum norm in an adaptive way, for any Hölder regularity of the true density between 0 and 1, thereby providing adaptive counterparts to the results for classical Pólya trees in Castillo (Ann. Inst. Henri Poincaré Probab. Stat. 53 (2017) 2074–2102). Second, the question of uncertainty quantification is considered. An adaptive nonparametric Bernstein–von Mises theorem is derived. Next, it is shown that, under a self-similarity condition on the true density, certain credible sets from the posterior distribution are adaptive confidence bands, having prescribed coverage level and with a diameter shrinking at optimal rate in the minimax sense.

Dans le modèle d’estimation de densité, la question de l’inférence adaptative est considérée, au moyen de lois a priori de type arbres de Pólya. Une classe de lois a priori à structure d’arbre, appelée “spike-and-slab Pólya trees”, est introduite. Des résultats de deux types sont obtenus. D’une part, il est établi que la loi a posteriori bayésienne converge à vitesse minimax en terme de la norme-sup de façon adaptative pour toute régularité de Hölder entre 0 et 1, obtenant ainsi des versions adaptatives des résultats pour les arbres de Pólya classiques de Castillo (Ann. Inst. Henri Poincaré Probab. Stat. 53 (2017) 2074–2102). D’autre part, nous considérons le problème de la quantification de l’incertitude. Un théorème de Bernstein–von Mises nonparamétrique et adaptatif est obtenu. Puis, il est établi que, sous conditions d’auto-similarité, des régions de crédibilité bien choisies de la loi a posteriori forment des bandes de confiance adaptatives, de niveau de confiance pré-déterminé, et de diamètre de taille optimale au sens minimax.

Acknowledgements

The authors would like to thank two referees as well as Thibault Randrianarisoa for insightful comments and suggestions on the paper.

Citation

Download Citation

Ismaël Castillo. Romain Mismer. "Spike and slab Pólya tree posterior densities: Adaptive inference." Ann. Inst. H. Poincaré Probab. Statist. 57 (3) 1521 - 1548, August 2021. https://doi.org/10.1214/20-AIHP1132

Information

Received: 30 November 2019; Revised: 17 November 2020; Accepted: 17 November 2020; Published: August 2021
First available in Project Euclid: 22 July 2021

MathSciNet: MR4291462
zbMATH: 1493.62240
Digital Object Identifier: 10.1214/20-AIHP1132

Subjects:
Primary: 62G20
Secondary: 62G07 , 62G15

Keywords: Bayesian nonparametrics , Bernstein–von Mises theorem , hierarchical Bayes , Pólya trees , Spike-and-slab priors , Supremum norm convergence

Rights: Copyright © 2021 Association des Publications de l’Institut Henri Poincaré

JOURNAL ARTICLE
28 PAGES

This article is only available to subscribers.
It is not available for individual sale.
+ SAVE TO MY LIBRARY

Vol.57 • No. 3 • August 2021
Back to Top