References & Citations
Mathematics > Statistics Theory
Title: Global identifiability of linear structural equation models
(Submitted on 4 Mar 2010 (v1), last revised 13 May 2011 (this version, v3))
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.
Submission history
From: Mathias Drton [view email][v1] Thu, 4 Mar 2010 21:53:45 GMT (30kb)
[v2] Fri, 17 Sep 2010 23:15:07 GMT (35kb)
[v3] Fri, 13 May 2011 09:15:00 GMT (199kb)
Link back to: arXiv, form interface, contact.