Abstract:Distributed estimation methods have recently been used to compute the maximum likelihood estimate of the precision matrix for large graphical Gaussian models. Our aim, in this article, is to give a Bayesian estimate of the precision matrix for large graphical Gaussian models with, additionally, symmetry constraints imposed by an underlying graph which is coloured. We take the sample posterior mean of the precision matrix as our estimate. We study its asymptotic behaviour under the regular asymptotic regime when the number of variables p is fixed and under the double asymptotic regime when both p and n grow to infinity. We show in particular that when the number of parameters of the local models is uniformly bounded the standard convergence rate of our estimate of the precision matrix to its true value, in the Frobenius norm, compares well with the rates in the current literature for the maximum likelihood estimate. The Canadian Journal of Statistics 46: 176–203; 2018 © 2017 Statistical Society of Canada Des méthodes distribuées d'estimation ont récemment été utilisées pour calculer l'estimateur au maximum de vraisemblance de la matrice de précision pour de grands modèles graphiques gaussiens. Les auteures proposent un estimateur bayésien de cette matrice dans le cas où une contrainte de symétrie est imposée par un graphique coloré sous‐jacent. Elles utilisent comme estimateur la moyenne a posteriori de la matrice de précision dont elles étudient le comportement asymptotique sous le régime habituel où le nombre de variables p est fixe, mais également sous le régime double où n et p tendent tous les deux vers l'infini. Lorsque le nombre de paramètres des modèles locaux est uniformément borné, les auteures montrent que le taux de convergence dans la norme de Frobenius de leur estimateur vers sa vraie valeur se compare avantageusement aux taux pour l'estimateur au maximum de vraisemblance trouvés dans la littérature courante. La revue canadienne de statistique 46: 176–203; 2018 © 2017 Société statistique du Canada

Precision matrix estimation under the horseshoe-like prior–penalty dual

Concentration of a Sparse Bayesian Model With Horseshoe Prior in Estimating High‐Dimensional Precision Matrix

Concentration of a sparse Bayesian model with Horseshoe prior in estimating high-dimensional precision matrix

Gaussian Graphical Model Estimation and Selection for High-Dimensional Incomplete Data Using Multiple Imputation and Horseshoe Estimators

Bayesian estimation for longitudinal data in a joint model with HPCs

Estimation of multiple networks in Gaussian mixture models.

A Data Driven Bayesian Graphical Ridge Estimator

Covariance Structure Estimation with Laplace Approximation

Approximate Bayesian estimation in large coloured graphical Gaussian models

Semi-parametric Expert Bayesian Network Learning with Gaussian Processes and Horseshoe Priors

The Effect of the Prior and the Experimental Design on the Inference of the Precision Matrix in Gaussian Chain Graph Models

Sparse reconstructions from few noisy data: analysis of hierarchical Bayesian models with generalized gamma hyperpriors

L1-Penalized Censored Gaussian Graphical Model

Horseshoe priors for edge-preserving linear Bayesian inversion

Maximum a posteriori estimation in graphical models using local linear approximation

Application of fused graphical lasso to statistical inference for multiple sparse precision matrices

Low-rank Bayesian matrix completion via geodesic Hamiltonian Monte Carlo on Stiefel manifolds

Targeted Fused Ridge Estimation of Inverse Covariance Matrices from Multiple High-Dimensional Data Classes

Efficient Distributed Estimation of High-Dimensional Sparse Precision Matrix

Generalized Precision Matrix for Scalable Estimation of Nonparametric Markov Networks

Penalty parameter selection and asymmetry corrections to Laplace approximations in Bayesian P-splines models