On the contraction properties of some high-dimensional quasi-posterior distributions

DOI10.1214/16-AOS1526MaRDI QIDQ1687130zbMATH OpenOpenAlexFDO

Authors Yves A. Atchadé

Publication date 22 December 2017

Published in The Annals of Statistics (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/1508.07929

zbMATH Keywords

binary graphical modelslogistic regression modelsquasi-Bayesian inferencequasi-likelihood functionsparsity-inducing prior

Mathematics Subject Classification ID

Bayesian inference (62F15) Multivariate distribution of statistics (62H10) Generalized linear models (logistic models) (62J12)

Abstract: We study the contraction properties of a quasi-posterior distribution $checkP{i}_{n,d}$ obtained by combining a quasi-likelihood function and a sparsity inducing prior distribution on $se{t}^d$, as both $n$ (the sample size), and $d$ (the dimension of the parameter) increase. We derive some general results that highlight a set of sufficient conditions under which $checkP{i}_{n,d}$ puts increasingly high probability on sparse subsets of $se{t}^d$, and contracts towards the true value of the parameter. We apply these results to the analysis of logistic regression models, and binary graphical models, in high-dimensional settings. For the logistic regression model, we shows that for well-behaved design matrices, the posterior distribution contracts at the rate $O(sqrt{s}_{s}tarlog(d)/n)$, where $s_{s}tar$ is the number of non-zero components of the parameter. For the binary graphical model, under some regularity conditions, we show that a quasi-posterior analog of the neighborhood selection of cite{meinshausen06} contracts in the Frobenius norm at the rate $O(sqrt(p+S)log(p)/n)$, where $p$ is the number of nodes, and $S$ the number of edges of the true graph.

Review text

This article studies the contraction properties of quasi-posterior distributions generated combining some types of quasi-likelihood functions and of sparsity inducing prior distributions on \(\mathbb R^d\). Sufficient conditions for the quasi-posterior distribution placing increasingly high probability on sparse subsets of \(\mathbb R^d\), and contracting toward the true value of the parameter as the sample size and the dimension of the parameter space increase, are established. Applications to high-dimensional logistic regression models and to high-dimensional binary graphical models are developed.

Recommendations

• Contraction properties of shrinkage priors in logistic regression
• Finite sample posterior concentration in high-dimensional regression
• Posterior consistency in linear models under shrinkage priors
• High dimensional posterior convergence rates for decomposable graphical models
• Quasi-Bayesian estimation of large Gaussian graphical models

Cited in

(13)

• Robust Bayes-like estimation: rho-Bayes estimation
• Unified Bayesian theory of sparse linear regression with nuisance parameters
• Posterior contraction in group sparse logit models for categorical responses
• Contraction of a quasi-Bayesian model with shrinkage priors in precision matrix estimation
• Quasi-Bayesian estimation of large Gaussian graphical models
• On frequentist coverage errors of Bayesian credible sets in moderately high dimensions
• Contraction properties of shrinkage priors in logistic regression
• Horseshoe Regularisation for Machine Learning in Complex and Deep Models¹
• From robust tests to Bayes-like posterior distributions
• On high-dimensional classification by sparse generalized Bayesian logistic regression
• Bayesian inference in high-dimensional models
• Spatiotemporal Besov Priors for Bayesian Inverse Problems
• Bayesian group regularization in generalized linear models with a continuous spike-and-slab prior

This page was built for publication: On the contraction properties of some high-dimensional quasi-posterior distributions

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q1687130)