Global identifiability of linear structural equation models

Abstract

Structural equation models are multivariate statistical models that are defined by specifying noisy functional relationships among random variables. We consider the classical case of linear relationships and additive Gaussian noise terms. We give a necessary and sufficient condition for global identifiability of the model in terms of a mixed graph encoding the linear structural equations and the correlation structure of the error terms. Global identifiability is understood to mean injectivity of the parametrization of the model and is fundamental in particular for applicability of standard statistical methodology.

Published in at http://dx.doi.org/10.1214/10-AOS859 the Annals of Statistics (http://www.imstat.org/aos/) by the Institute of Mathematical Statistics (http://www.imstat.org)

References in corpus (3)

• Likelihood ratio tests and singularities
• Computing maximum likelihood estimates in recursive linear models with correlated errors
• Probability distributions with summary graph structure

Cited by in corpus (18)

• Half-trek criterion for generic identifiability of linear structural equation models
• Differentiable Causal Discovery Under Unmeasured Confounding
• Identification of discrete concentration graph models with one hidden binary variable
• Counterfactual analyses with graphical models based on local independence
• Efficient least squares for estimating total effects under linearity and causal sufficiency
• Complexity analysis of Bayesian learning of high-dimensional DAG models and their equivalence classes
• Identifiability of directed Gaussian graphical models with one latent source
• Stability of Linear Structural Equation Models of Causal Inference
• Causal Inference for Influence Propagation -- Identifiability of the Independent Cascade Model
• A Topological Perspective on Causal Inference
• On the global identifiability of logistic regression models with misclassified outcomes
• Structural Identifiability of Cyclic Graphical Models of Biological Networks with Latent Variables
• Learning linear acyclic causal model including Gaussian noise using ancestral relationships
• Generic Identifiability of Linear Structural Equation Models by Ancestor Decomposition
• Non-Identifiability in Network Autoregressions
• Dependency in DAG models with Hidden Variables
• Algebraic Properties of Gaussian HTC-identifiable Graphs
• Robust Identifiability in Linear Structural Equation Models of Causal Inference