On the toric algebra of graphical models

Abstract

We formulate necessary and sufficient conditions for an arbitrary discrete probability distribution to factor according to an undirected graphical model, or a log-linear model, or other more general exponential models. For decomposable graphical models these conditions are equivalent to a set of conditional independence statements similar to the Hammersley--Clifford theorem; however, we show that for nondecomposable graphical models they are not. We also show that nondecomposable models can have nonrational maximum likelihood estimates. These results are used to give several novel characterizations of decomposable graphical models.

Published at http://dx.doi.org/10.1214/009053606000000263 in the Annals of Statistics (http://www.imstat.org/aos/) by the Institute of Mathematical Statistics (http://www.imstat.org)

Cited by in corpus (62)

• Maximum likelihood estimation in log-linear models
• Conjunctive Bayesian networks
• Maximum lilkelihood estimation in the $\beta$-model
• The Maximum Likelihood Degree of Toric Varieties
• Quantum Correlations in the Minimal Scenario
• Implicit inequality constraints in a binary tree model
• Support Sets in Exponential Families and Oriented Matroid Theory
• Total positivity in exponential families with application to binary variables
• Neighborliness of Marginal Polytopes
• Markov bases for two-way subtable sum problems
• Decompositions of Binomial Ideals
• Maximizing the divergence from a hierarchical model of quantum states
• A Markov Basis for Conditional Test of Common Diagonal Effect in Quasi-Independence Model for Square Contingency Tables
• Indispensable binomials in semigroup ideals
• Algebraic statistical models
• Minimal and minimal invariant Markov bases of decomposable models for contingency tables
• Equations defining probability tree models
• Boundary cliques, clique trees and perfect sequences of maximal cliques of a chordal graph
• Three Counterexamples on Semigraphoids
• Logarithmic Voronoi cells
• Positive margins and primary decomposition
• Gröbner bases for staged trees
• Computing images of polynomial maps
• Staged tree models with toric structure
• Toric invariant theory for maximum likelihood estimation in log-linear models
• Algebraic Bayesian analysis of contingency tables with possibly zero-probability cells
• A note on the border of an exponential family
• On the Existence of the MLE for a Directed Random Graph Network Model with Reciprocation
• Betti numbers of Stanley-Reisner rings determine hierarchical Markov degrees
• Relations among conditional probabilities
• Rational partition models under iterative proportional scaling
• A Gröbner basis characterization for chordal comparability graphs
• Symmetric measures via moments
• Potential unsatisfiability of cyclic constraints on stochastic biological networks biases selection toward hierarchical architectures
• Quantum marginals, faces, and coatoms
• Minimal primes of ideals arising from conditional independence statements
• A variation principle for ground spaces
• Algebraic causality: Bayes nets and beyond
• Nonlinear Algebra and Applications
• Graphical models for correlated defaults
• Sampling lattice points in a polytope: a Bayesian biased algorithm with random updates
• Toric and non-toric Bayesian networks
• Markov bases of binary graph models of K_4-minor free graphs
• Mixture decompositions of exponential families using a decomposition of their sample spaces
• Ideals of Graph Homomorphisms
• Hierarchical Models, Marginal Polytopes, and Linear Codes
• Mixtures of Discrete Decomposable Graphical Models
• On the Geometry of Discrete Exponential Families with Application to Exponential Random Graph Models
• Graphical models in Macaulay2
• Normal binary graph models
• Graphical methods for efficient likelihood inference in Gaussian covariance models
• On the closure of relational models
• Factorization of Discrete Probability Distributions
• Specializations of multigradings and the arithmetical rank of lattice ideals
• Polynomial relations among principal minors of a 4x4-matrix
• A class of statistical models to weaken independence in two-way contingency tables
• Maximizing the Bregman divergence from a Bregman family
• On decomposable and reducible integer matrices
• Optimally approximating exponential families
• Completions to Discrete Probability Distributions in Log-linear Models
• Maximum likelihood estimation of log-affine models using detailed-balanced reaction networks
• Discovery of statistical equivalence classes using computer algebra