Markovian acyclic directed mixed graphs for discrete data

Abstract

Acyclic directed mixed graphs (ADMGs) are graphs that contain directed ($\to$) and bidirected ($\leftrightarrow$) edges, subject to the constraint that there are no cycles of directed edges. Such graphs may be used to represent the conditional independence structure induced by a DAG model containing hidden variables on its observed margin. The Markovian model associated with an ADMG is simply the set of distributions obeying the global Markov property, given via a simple path criterion (m-separation). We first present a factorization criterion characterizing the Markovian model that generalizes the well-known recursive factorization for DAGs. For the case of finite discrete random variables, we also provide a parameterization of the model in terms of simple conditional probabilities, and characterize its variation dependence. We show that the induced models are smooth. Consequently, Markovian ADMG models for discrete variables are curved exponential families of distributions.

Published in at http://dx.doi.org/10.1214/14-AOS1206 the Annals of Statistics (http://www.imstat.org/aos/) by the Institute of Mathematical Statistics (http://www.imstat.org)

References in corpus (9)

• Probabilistic Evaluation of Sequential Plans from Causal Models with Hidden Variables
• Identifying the consequences of dynamic treatment strategies: A decision-theoretic overview
• Discrete chain graph models
• Markov properties for mixed graphs
• Identification of Conditional Interventional Distributions
• Probability distributions with summary graph structure
• Stable mixed graphs
• A factorization criterion for acyclic directed mixed graphs
• Maximum likelihood fitting of acyclic directed mixed graphs to binary data

Cited by in corpus (21)

• Graphs for margins of Bayesian networks
• Margins of discrete Bayesian networks
• Markov Properties for Graphical Models with Cycles and Latent Variables
• Markov equivalence of marginalized local independence graphs
• Causal Etiology of the Research of James M. Robins
• Differentiable Causal Discovery Under Unmeasured Confounding
• Marginalization and Conditioning for LWF Chain Graphs
• Quantum Causal Models
• Estimating causal structure using conditional DAG models
• Efficient adjustment sets for population average treatment effect estimation in non-parametric causal graphical models
• CADET: Debugging and Fixing Misconfigurations using Counterfactual Reasoning
• Integer Programming for Causal Structure Learning in the Presence of Latent Variables
• Necessary and sufficient graphical conditions for optimal adjustment sets in causal graphical models with hidden variables
• On the Properties of MVR Chain Graphs
• Structural Learning of Multivariate Regression Chain Graphs via Decomposition
• Learning latent causal graphs via mixture oracles
• Smooth, identifiable supermodels of discrete DAG models with latent variables
• Confounder selection via iterative graph expansion
• Faster algorithms for Markov equivalence
• The neighborhood lattice for encoding partial correlations in a Hilbert space
• Path Dependent Structural Equation Models