High-dimensional multivariate posterior consistency under global-local shrinkage priors

DOI10.1016/J.JMVA.2018.04.010zbMATH Open1403.62134arXiv1711.07635OpenAlexW2768228643WikidataQ129915462 ScholiaQ129915462MaRDI QIDQ1661340FDOQ1661340

Malay Ghosh, Ray Bai

Publication date: 16 August 2018

Published in: Journal of Multivariate Analysis (Search for Journal in Brave)

Abstract: We consider sparse Bayesian estimation in the classical multivariate linear regression model with $p$ regressors and $q$ response variables. In univariate Bayesian linear regression with a single response $y$, shrinkage priors which can be expressed as scale mixtures of normal densities are popular for obtaining sparse estimates of the coefficients. In this paper, we extend the use of these priors to the multivariate case to estimate a $pimesq$ coefficients matrix $mathbfB$. We derive sufficient conditions for posterior consistency under the Bayesian multivariate linear regression framework and prove that our method achieves posterior consistency even when $p>n$ and even when $p$ grows at nearly exponential rate with the sample size. We derive an efficient Gibbs sampling algorithm and provide the implementation in a comprehensive R package called MBSP. Finally, we demonstrate through simulations and data analysis that our model has excellent finite sample performance.

Full work available at URL: https://arxiv.org/abs/1711.07635

Recommendations

• Ultra high-dimensional multivariate posterior contraction rate under shrinkage priors
• Posterior consistency in linear models under shrinkage priors
• Finite sample posterior concentration in high-dimensional regression
• Bayesian linear regression for multivariate responses under group sparsity
• High-dimensional posterior consistency of the Bayesian Lasso

zbMATH Keywords

variable selectionhigh-dimensional datasparsityheavy tailposterior consistencyshrinkage estimation

Mathematics Subject Classification ID

Asymptotic properties of parametric estimators (62F12) Bayesian inference (62F15) Estimation in multivariate analysis (62H12) Ridge regression; shrinkage estimators (Lasso) (62J07)

Cites Work

• Sure independence screening in generalized linear models with NP-dimensionality
• The Adaptive Lasso and Its Oracle Properties
• Title not available (Why is that?)
• The Dantzig selector: statistical estimation when \(p\) is much larger than \(n\). (With discussions and rejoinder).
• The horseshoe estimator for sparse signals
• The Bayesian Lasso
• Sure Independence Screening for Ultrahigh Dimensional Feature Space
• Regularization and Variable Selection Via the Elastic Net
• Model Selection and Estimation in Regression with Grouped Variables
• The horseshoe+ estimator of ultra-sparse signals
• On the half-Cauchy prior for a global scale parameter
• Scaled sparse linear regression
• Title not available (Why is that?)
• Bayesian Variable Selection in Linear Regression
• Bayesian Generalized Low Rank Regression Models for Neuroimaging Phenotypes and Genetic Markers
• Some priors for sparse regression modelling
• Sparse Reduced-Rank Regression for Simultaneous Dimension Reduction and Variable Selection
• Joint variable and rank selection for parsimonious estimation of high-dimensional matrices
• Bayesian variable selection with shrinking and diffusing priors
• Spike and slab variable selection: frequentist and Bayesian strategies
• Multivariate Bayesian Variable Selection and Prediction
• Title not available (Why is that?)
• Sparse Partial Least Squares Regression for Simultaneous Dimension Reduction and Variable Selection
• The horseshoe estimator: posterior concentration around nearly black vectors
• Bayesian variable selection and estimation for group Lasso
• Conditions for posterior contraction in the sparse normal means problem
• Joint high-dimensional Bayesian variable and covariance selection with an application to eQTL analysis
• Posterior consistency in linear models under shrinkage priors
• Generalized double Pareto shrinkage
• Bayesian linear regression with sparse priors
• Asymptotic properties of Bayes risk for the horseshoe prior
• Proper Bayes Minimax Estimators of the Multivariate Normal Mean
• A robust generalized Bayes estimator and confidence region for a multivariate normal mean
• Asymptotic properties of Bayes risk of a general class of shrinkage priors in multiple hypothesis testing under sparsity
• Multivariate sparse group lasso for the multivariate multiple linear regression with an arbitrary group structure
• Uncertainty quantification for the horseshoe (with discussion)
• Bayesian variable selection regression of multivariate responses for group data
• Asymptotic optimality of one-group shrinkage priors in sparse high-dimensional problems
• Bayesian sparse reduced rank multivariate regression
• Adaptive posterior contraction rates for the horseshoe
• An algorithm for the multivariate group lasso with covariance estimation

Cited In (8)

• The EAS approach to variable selection for multivariate response data in high-dimensional settings
• A Comparison of Bayesian Multivariate Versus Univariate Normal Regression Models for Prediction
• Summaries of three keynote lectures at the SAE – 2018
• MBSP
• High-dimensional Bayesian network classification with network global-local shrinkage priors
• Ultra high-dimensional multivariate posterior contraction rate under shrinkage priors
• Statistical inference via conditional Bayesian posteriors in high-dimensional linear regression
• Bayesian sparse spiked covariance model with a continuous matrix shrinkage prior

Uses Software

• R
• glmnet
• R2GUESS

This page was built for publication: High-dimensional multivariate posterior consistency under global-local shrinkage priors

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q1661340)