Bayesian graph selection consistency under model misspecification

DOI10.3150/20-BEJ1253zbMATH Open1476.62121arXiv1901.04134MaRDI QIDQ2214264FDOQ2214264

Authors: Yabo Niu, Debdeep Pati, Bani Mallick 

Publication date: 7 December 2020

Published in: Bernoulli (Search for Journal in Brave)

Abstract: Gaussian graphical models are a popular tool to learn the dependence structure in the form of a graph among variables of interest. Bayesian methods have gained in popularity in the last two decades due to their ability to simultaneously learn the covariance and the graph and characterize uncertainty in the selection. For scalability of the Markov chain Monte Carlo algorithms, decomposability is commonly imposed on the graph space. A wide variety of graphical conjugate priors are proposed jointly on the covariance matrix and the graph with improved algorithms to search along the space of decomposable graphs, rendering the methods extremely popular in the context of multivariate dependence modeling. {it An open problem} in Bayesian decomposable structure learning is whether the posterior distribution is able to select a meaningful decomposable graph that it is ``close in an appropriate sense to the true non-decomposable graph, when the dimension of the variables increases with the sample size. In this article, we explore specific conditions on the true precision matrix and the graph which results in an affirmative answer to this question using a commonly used hyper-inverse Wishart prior on the covariance matrix and a suitable complexity prior on the graph space, both in the well-specified and misspecified settings. In absence of structural sparsity assumptions, our strong selection consistency holds in a high dimensional setting where $p=O(n^{alpha})$ for $alpha<1/3$. We show when the true graph is non-decomposable, the posterior distribution on the graph concentrates on a set of graphs that are {it minimal triangulations} of the true graph.

Full work available at URL: https://arxiv.org/abs/1901.04134

Recommendations

• Bayesian inference for high-dimensional decomposable graphs
• The performance of covariance selection methods that consider decomposable models only
• Bayesian structure learning in graphical models
• Posterior graph selection and estimation consistency for high-dimensional Bayesian DAG models
• scientific article; zbMATH DE number 5769726

zbMATH Keywords

Gaussian graphical modelmodel misspecificationpartial correlationminimal triangulationdecomposable graphgraph selection consistencyhyper-inverse Wishart distribution

Mathematics Subject Classification ID

Bayesian inference (62F15) Applications of graph theory (05C90) Probabilistic graphical models (62H22) Algebraic statistics (62R01)

Cites Work

• Sparse graphical models for exploring gene expression data
• Experiments in stochastic computation for high-dimensional graphical models
• Reversible jump Markov chain Monte Carlo computation and Bayesian model determination
• High-dimensional graphs and variable selection with the Lasso
• Multiple testing and error control in Gaussian graphical model selection
• Sparsistency and rates of convergence in large covariance matrix estimation
• Conjugate priors for exponential families
• Hyper Markov laws in the statistical analysis of decomposable graphical models
• Regularized estimation of large covariance matrices
• Model selection and estimation in the Gaussian graphical model
• Title not available (Why is that?)
• Operator norm consistent estimation of large-dimensional sparse covariance matrices
• Wishart distributions for decomposable graphs
• Adaptive thresholding for sparse covariance matrix estimation
• Cholesky decomposition of a hyper inverse Wishart matrix
• Simulation of hyper-inverse Wishart distributions in graphical models
• Flexible covariance estimation in graphical Gaussian models
• Objective Bayesian model selection in Gaussian graphical models
• Hyper Inverse Wishart Distribution for Non-decomposable Graphs and its Application to Bayesian Inference for Gaussian Graphical Models
• Decomposable graphical Gaussian model determination
• A Monte Carlo method for computing the marginal likelihood in nondecomposable Gaussian graphical models
• Minimal triangulations of graphs: a survey
• Algorithmic Aspects of Vertex Elimination on Graphs
• Posterior convergence rates for estimating large precision matrices using graphical models
• Title not available (Why is that?)
• The performance of covariance selection methods that consider decomposable models only
• Bayesian Inference for Gaussian Graphical Models Beyond Decomposable Graphs
• Simulation of hyper-inverse Wishart distributions for non-decomposable graphs
• An empirical Bayes procedure for the selection of Gaussian graphical models
• Posterior graph selection and estimation consistency for high-dimensional Bayesian DAG models
• Exact formulas for the normalizing constants of Wishart distributions for graphical models
• Bayesian structure learning in graphical models
• Minimax posterior convergence rates and model selection consistency in high-dimensional DAG models based on sparse Cholesky factors

Cited In (11)

• Compatible priors for model selection of high-dimensional Gaussian DAGs
• An empirical Bayes procedure for the selection of Gaussian graphical models
• The cost of using decomposable Gaussian graphical models for computational convenience
• On the non-local priors for sparsity selection in high-dimensional Gaussian DAG models
• The G-Wishart Weighted Proposal Algorithm: Efficient Posterior Computation for Gaussian Graphical Models
• Robust concentration graph model selection
• The Missing Consistency Theorem for Bayesian Learning: Stochastic Model Selection
• Bayesian inference for high-dimensional decomposable graphs
• On a wider class of prior distributions for graphical models
• The performance of covariance selection methods that consider decomposable models only
• Bayesian learning of graph substructures

Uses Software

• HdBCS

This page was built for publication: Bayesian graph selection consistency under model misspecification

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2214264)