On generating random Gaussian graphical models
arXiv:1909.01062 · doi:10.1016/j.ijar.2020.07.007
Abstract
Structure learning methods for covariance and concentration graphs are often validated on synthetic models, usually obtained by randomly generating: (i) an undirected graph, and (ii) a compatible symmetric positive definite (SPD) matrix. In order to ensure positive definiteness in (ii), a dominant diagonal is usually imposed. In this work we investigate different methods to generate random symmetric positive definite matrices with undirected graphical constraints. We show that if the graph is chordal it is possible to sample uniformly from the set of correlation matrices compatible with the graph, while for general undirected graphs we rely on a partial orthogonalization method.
arXiv commentImproved figures, algorithm descriptions and text exposition. arXiv admin note: substantial text overlap with arXiv:1807.03090
References in corpus (10)
- Nonlinear shrinkage estimation of large-dimensional covariance matrices
- Regularized estimation of large-scale gene association networks using graphical Gaussian models
- Regularized rank-based estimation of high-dimensional nonparanormal graphical models
- Wishart distributions for decomposable graphs
- Wishart distributions for decomposable covariance graph models
- Simplicial faces of the set of correlation matrices
- Bayesian Estimation of Gaussian Graphical Models with Predictive Covariance Selection
- A fast Metropolis-Hastings method for generating random correlation matrices
- A partial orthogonalization method for simulating covariance and concentration graph matrices
- Markov Network Structure Learning via Ensemble-of-Forests Models