Abstract: Directed acyclic graph (DAG) models, also called Bayesian networks, impose conditional independence constraints on a multivariate probability distribution, and are widely used in probabilistic reasoning, machine learning and causal inference. If latent variables are included in such a model, then the set of possible marginal distributions over the remaining (observed) variables is generally complex, and not represented by any DAG. Larger classes of mixed graphical models, which use multiple edge types, have been introduced to overcome this; however, these classes do not represent all the models which can arise as margins of DAGs. In this paper we show that this is because ordinary mixed graphs are fundamentally insufficiently rich to capture the variety of marginal models. We introduce a new class of hyper-graphs, called mDAGs, and a latent projection operation to obtain an mDAG from the margin of a DAG. We show that each distinct marginal of a DAG model is represented by at least one mDAG, and provide graphical results towards characterizing when two such marginal models are the same. Finally we show that mDAGs correctly capture the marginal structure of causally-interpreted DAGs under interventions on the observed variables.
Recommendations
Cites work
- A new approach to causal inference in mortality studies with a sustained exposure period—application to control of the healthy worker survivor effect
- Ancestral graph Markov models.
- Causality. Models, reasoning, and inference
- Comments on: Sequences of regressions and their independencies
- Factor graphs and the sum-product algorithm
- Half-trek criterion for generic identifiability of linear structural equation models
- Independence properties of directed markov fields
- Likelihood ratio tests and singularities
- Marginalizing and conditioning in graphical models
- Margins of discrete Bayesian networks
- Markov Properties for Acyclic Directed Mixed Graphs
- Markovian acyclic directed mixed graphs for discrete data
- Mendelian randomization as an instrumental variable approach to causal inference
- Modeling and Reasoning with Bayesian Networks
- On Dedekind's Problem: The Number of Monotone Boolean Functions
- On the causal interpretation of acyclic mixed graphs under multivariate normality
- Probability distributions with summary graph structure
- Statistics, causality and Bell's theorem
- Tensors of nonnegative rank two
Cited in
(23)- Nested Markov properties for acyclic directed mixed graphs
- Marginals of DAG-Isomorphic Independence Models
- Discovering agents
- Identifying collapsible sets in directed graphical models via inducing paths
- Bayesian networks: the minimal triangulations of a graph
- scientific article; zbMATH DE number 7306884 (Why is no real title available?)
- Probability based independence sampler for Bayesian quantitative learning in graphical log-linear marginal models
- DAGs as minimal \(\mathcal{I} \)-maps for the induced models of causal Bayesian networks under conditioning
- Analysing causal structures with entropy
- Margins of discrete Bayesian networks
- On a hypergraph probabilistic graphical model
- Marginalizing and conditioning in graphical models
- Thinning a Triangulation of a Bayesian Network or Undirected Graph to Create a Minimal Triangulation
- A convergent inflation hierarchy for quantum causal structures
- Learning Moral Graphs in Construction of High-Dimensional Bayesian Networks for Mixed Data
- Latent free equivalent mDAGs
- An Automated Approach to Causal Inference in Discrete Settings
- Marginal Markov subgraph of Bayesian network and its applications
- Foundations of structural causal models with cycles and latent variables
- Markov equivalence of marginalized local independence graphs
- The hidden life of latent variables: Bayesian learning with mixed graph models
- Causal structure learning: a combinatorial perspective
- The marginal factorization of Bayesian networks and its application
This page was built for publication: Graphs for margins of Bayesian networks
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2821469)
