Parameter priors for directed acyclic graphical models and the characterization of several probability distributions

DOI10.1214/AOS/1035844981MaRDI QIDQ79791zbMATH OpenOpenAlexFDO

Authors Dan Geiger, David Heckerman

Publication date 1 October 2002

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

Full work available at URL https://arxiv.org/abs/2105.03248

zbMATH Keywords

Bayesian networkdirected acyclic graphical modelDirichlet distributionGaussian DAG modellearninglinear regression modelnormal distributionWishart distribution

Mathematics Subject Classification ID

Directed graphs (digraphs), tournaments (05C20) Applications of graph theory (05C90) Functional equations and inequalities (39B99) Probability distributions: general theory (60E05) Bayesian problems; characterization of Bayes procedures (62C10) Characterization and structure theory for multivariate probability distributions; copulas (62H05)

Abstract: We develop simple methods for constructing parameter priors for model choice among Directed Acyclic Graphical (DAG) models. In particular, we introduce several assumptions that permit the construction of parameter priors for a large number of DAG models from a small set of assessments. We then present a method for directly computing the marginal likelihood of every DAG model given a random sample with no missing observations. We apply this methodology to Gaussian DAG models which consist of a recursive set of linear regression models. We show that the only parameter prior for complete Gaussian DAG models that satisfies our assumptions is the normal-Wishart distribution. Our analysis is based on the following new characterization of the Wishart distribution: let $W$ be an $nimesn$, $nge3$, positive-definite symmetric matrix of random variables and $f(W)$ be a pdf of $W$. Then, f$(W)$ is a Wishart distribution if and only if $W_{11}-W_{12}{W}_{22}^{-1}W{\text{'}}_{12}$ is independent of $W_{12},W_{22}$ for every block partitioning $W_{11},W_{12},W{\text{'}}_{12},W_{22}$ of $W$. Similar characterizations of the normal and normal-Wishart distributions are provided as well.

Recommendations

• Compatible priors for model selection of high-dimensional Gaussian DAGs
• Wishart distributions for decomposable graphs
• Prior distributions for Gaussian graphical models: a comparison between the directed and undirected case
• Exact formulas for the normalizing constants of Wishart distributions for graphical models
• Wishart distributions for decomposable covariance graph models

Cites work

• A Bayesian method for the induction of probabilistic networks from data
• A characterization of Markov equivalence classes for acyclic digraphs
• A characterization of the bivariate Wishart distribution
• A characterization of the Dirichlet distribution through global and local parameter independence
• Bayesian analysis in expert systems. With comments and a rejoinder by the authors
• Bayesian model averaging and model selection for markov equivalence classes of acyclic digraphs
• scientific article; zbMATH DE number 48812 (Why is no real title available?)
• scientific article; zbMATH DE number 578421 (Why is no real title available?)
• scientific article; zbMATH DE number 1522721 (Why is no real title available?)
• scientific article; zbMATH DE number 3426633 (Why is no real title available?)
• scientific article; zbMATH DE number 3225653 (Why is no real title available?)
• scientific article; zbMATH DE number 3357742 (Why is no real title available?)
• Hyper Markov laws in the statistical analysis of decomposable graphical models
• Learning Bayesian networks: The combination of knowledge and statistical data
• On regular solutions of functional equations
• Parameter priors for directed acyclic graphical models and the characterization of several probability distributions
• Regularity property of the functional equation of the Dirichlet distribution
• Sequential updating of conditional probabilities on directed graphical structures

Cited in

(63)

• Kshirsagar-Tan independence property of beta matrices and related characterizations
• Structural learning and estimation of joint causal effects among network-dependent variables
• A characterization of the Dirichlet distribution through global and local parameter independence
• Prior distributions for Gaussian graphical models: a comparison between the directed and undirected case
• Learning Markov equivalence classes of directed acyclic graphs: an objective Bayes approach
• Posterior graph selection and estimation consistency for high-dimensional Bayesian DAG models
• On the invariance of conditioning procedures for the specification of prior distributions for nested DAG models
• Sparse graphical models for exploring gene expression data
• Modeling risk contagion in the Italian zonal electricity market
• NetVIX -- a network volatility index of financial markets
• Bayesian graphical models for modern biological applications
• Compatible priors for model selection of high-dimensional Gaussian DAGs
• Equivalence class selection of categorical graphical models
• Objective Bayes model selection of Gaussian interventional essential graphs for the identification of signaling pathways
• A review of Gaussian Markov models for conditional independence
• Discovering causal graphs with cycles and latent confounders: an exact branch-and-bound approach
• Wishart distributions for decomposable covariance graph models
• The Matsumoto\,-\,Yor property and the structure of the Wishart distribution
• Wishart distributions for decomposable graphs
• A characterization of the bivariate Wishart distribution
• Distribution of the product of a singular Wishart matrix and a normal vector
• Bayesian approaches for large biological networks
• Estimating Bayesian networks for high-dimensional data with complex mean structure and random effects
• Objective Bayesian search of Gaussian directed acyclic graphical models for ordered variables with non-local priors
• A new prior for discrete DAG models with a restricted set of directions
• Objective Bayes factors for Gaussian directed acyclic graphical models
• Kummer and gamma laws through independences on trees -- another parallel with the Matsumoto-Yor property
• A Gibbs sampler for learning DAG: a unification for discrete and Gaussian domains
• On the exact and approximate distributions of the product of a Wishart matrix with a normal vector
• Sensitivity to hyperprior parameters in Gaussian Bayesian networks
• BiDAG
• Hyper Inverse Wishart Distribution for Non-decomposable Graphs and its Application to Bayesian Inference for Gaussian Graphical Models
• The use of unlabeled data in predictive modeling
• Addendum on the scoring of Gaussian directed acyclic graphical models
• Compatible Prior Distributions for Directed Acyclic Graph Models
• Parameter Estimation for Undirected Graphical Models With Hard Constraints
• scientific article; zbMATH DE number 7626781 (Why is no real title available?)
• Efficient Sampling and Structure Learning of Bayesian Networks
• Learning Gaussian graphical models with fractional marginal pseudo-likelihood
• Comparing score-based methods for estimating Bayesian networks using the Kullback-Leibler divergence
• Efficient local updates for undirected graphical models
• Bayesian inference of causal effects from observational data in Gaussian graphical models
• Causal structure learning: a combinatorial perspective
• A loss‐based prior for Gaussian graphical models
• Bayesian Model Selection of Gaussian Directed Acyclic Graph Structures
• Complexity analysis of Bayesian learning of high-dimensional DAG models and their equivalence classes
• Model averaging for sparse seemingly unrelated regression using Bayesian networks among the errors
• Being Bayesian about learning Gaussian Bayesian networks from incomplete data
• Bayesian causal inference in probit graphical models
• Being Bayesian about learning Bayesian networks from ordinal data
• Bayesian sample size determination for causal discovery
• High-dimensional covariance estimation for Gaussian directed acyclic graph models with given order
• Bayesian learning of multiple directed networks from observational data
• Bayesian graphical modeling for heterogeneous causal effects
• Continuity approximation in hybrid Bayesian networks structure learning
• Being Bayesian about learning Bayesian networks from hybrid data
• Learning Bayesian networks from ordinal data -- the Bayesian way
• Joint structure learning and causal effect estimation for categorical graphical models
• A BGe score for tied-covariance mixture Gaussian Bayesian networks
• More on connections between Wishart and matrix GIG distributions
• Parametrizations and reference priors for multinomial decomposable graphical models
• Parameter priors for directed acyclic graphical models and the characterization of several probability distributions
• On the robustness of Bayesian networks to learning from non-conjugate sampling

Describes a project that uses

Uses Software

• elicit-normlin

This page was built for publication: Parameter priors for directed acyclic graphical models and the characterization of several probability distributions

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q79791)