Nested covariance determinants and restricted trek separation in Gaussian graphical models. (English) Zbl 1476.62119

This paper is placed in the context of Gaussian graphical models [S. L. Lauritzen, Graphical models. Oxford: Oxford Univ. Press (1996; Zbl 0907.62001)], and its focus is on acyclic directed mixed graphs (ADMGs), which are an extension of the better known directed acylic graphs (DAGs). Specifically, the authors are interested in explaining the relations among the entries of a covariance matrix that belongs to such linear structural equation models.
It is well known in the graphical models literature that probabilistic conditional independence (CI) statements among the entries of Gaussian random vectors correspond to the vanishing of specific minors of the covariance matrix, i.e., a CI statement holds if and only if the determinant of certain submatrix of the covariance matrix is zero. These are enough to completely describe the constraints in the covariance matrix of a DAG model, and correspond graphically to d-separation [M. Studený, Probabilistic conditional independence structures. London: Springer (2005; Zbl 1070.62001)]. However, in the presence of hidden variables (represented in the mixed graph by bidirected edges between observed nodes) this is no longer the case. In particular, there may be minors that vanish on the model that do not correspond to CI statements.
The work of [S. Sullivant et al., Ann. Stat. 38, No. 3, 1665–1685 (2010; Zbl 1189.62091)] introduced the concept of trek separation and managed to characterize all vanishing minors of the covariance matrix. However, there may exist other polynomial constraints that hold for ADMG models that are not of this form, such as the (in)famous Verma constraint [T. van Ommen and J. M. Mooij, “Algebraic equivalence of linear structural equation models”, in: Proceedings of the 33rd Annual Conference on Uncertainty in Artificial Intelligence (UAI-17) (2017), arXiv:1807.03527]. These kinds of relations appear to be harder to understand, and the main contribution of the paper under review is that while they cannot be expressed as minors of the covariance matrix, they can actually be expressed as determinants of matrices whose entries are determinants themselves: nested determinants. Furthermore, in their main theorem the authors are able to explain the structure of many of these instances via the concept of restricted trek-separation.
Through several illuminating and well-chosen examples, the authors illustrate not only how the theory they develop applies to different kinds of ADMGs, but also how there are other graphical models, such as ones that now allow cycles, which have constraints that also admit a nested determinant representation (which is in general not unique). Indeed, these observations reveal that the story of nested determinants and graphical model relations is not fully understood yet, and the paper suggests many interesting open problems for follow-up work in this direction.


62H22 Probabilistic graphical models
62H12 Estimation in multivariate analysis
62J10 Analysis of variance and covariance (ANOVA)
62R01 Algebraic statistics
00A27 Lists of open problems


Full Text: DOI arXiv Euclid


[1] Chen, B. (2016). Identification and overidentification of linear structural equation models. In Advances in Neural Information Processing Systems 29 (D.D. Lee, M. Sugiyama, U.V. Luxburg, I. Guyon and R. Garnett, eds.) 1587-1595. Curran Associates.
[2] Draisma, J., Sullivant, S. and Talaska, K. (2013). Positivity for Gaussian graphical models. Adv. in Appl. Math. 50 661-674. · Zbl 1279.62135 · doi:10.1016/j.aam.2013.03.001
[3] Drton, M. (2009). Likelihood ratio tests and singularities. Ann. Statist. 37 979-1012. · Zbl 1196.62020 · doi:10.1214/07-AOS571
[4] Drton, M. (2018). Algebraic problems in structural equation modeling. In The 50th Anniversary of Gröbner Bases (T. Hibi, ed.). Adv. Stud. Pure Math. 77 35-86. Tokyo: Math. Soc. Japan. · Zbl 1411.62128
[5] Drton, M., Foygel, R. and Sullivant, S. (2011). Global identifiability of linear structural equation models. Ann. Statist. 39 865-886. · Zbl 1215.62052 · doi:10.1214/10-AOS859
[6] Drton, M. and Maathuis, M.H. (2017). Structure learning in graphical modeling. Annu. Rev. Stat. Appl. 4 365-393.
[7] Drton, M., Sturmfels, B. and Sullivant, S. (2007). Algebraic factor analysis: Tetrads, pentads and beyond. Probab. Theory Related Fields 138 463-493. · Zbl 1111.13020 · doi:10.1007/s00440-006-0033-2
[8] Drton, M., Sturmfels, B. and Sullivant, S. (2009). Lectures on Algebraic Statistics. Oberwolfach Seminars 39. Basel: Birkhäuser. · Zbl 1166.13001
[9] Drton, M. and Weihs, L. (2016). Generic identifiability of linear structural equation models by ancestor decomposition. Scand. J. Stat. 43 1035-1045. · Zbl 1373.62250 · doi:10.1111/sjos.12227
[10] Drton, M. and Yu, J. (2010). On a parametrization of positive semidefinite matrices with zeros. SIAM J. Matrix Anal. Appl. 31 2665-2680. · Zbl 1210.15036 · doi:10.1137/100783170
[11] Evans, R.J. and Richardson, T.S. (2014). Markovian acyclic directed mixed graphs for discrete data. Ann. Statist. 42 1452-1482. · Zbl 1302.62148 · doi:10.1214/14-AOS1206
[12] Fink, A., Rajchgot, J. and Sullivant, S. (2016). Matrix Schubert varieties and Gaussian conditional independence models. J. Algebraic Combin. 44 1009-1046. · Zbl 1411.14060 · doi:10.1007/s10801-016-0698-2
[13] Foygel, R., Draisma, J. and Drton, M. (2012). Half-trek criterion for generic identifiability of linear structural equation models. Ann. Statist. 40 1682-1713. · Zbl 1257.62059 · doi:10.1214/12-AOS1012
[14] Frydenberg, M. (1990). The chain graph Markov property. Scand. J. Stat. 17 333-353. · Zbl 0713.60013
[15] Koster, J.T.A. (2002). Marginalizing and conditioning in graphical models. Bernoulli 8 817-840. · Zbl 1011.60026
[16] Lauritzen, S.L. (1996). Graphical Models. Oxford Statistical Science Series 17. New York: The Clarendon Press. · Zbl 0907.62001
[17] Lnenicka, R. and Matúš, F. (2007). On Gaussian conditional independent structures. Kybernetika (Prague) 43 327-342. · Zbl 1144.60302
[18] Pearl, J. (2009). Causality: Models, Reasoning, and Inference, 2nd ed. Cambridge: Cambridge Univ. Press. · Zbl 1188.68291
[19] Richardson, T. and Spirtes, P. (2002). Ancestral graph Markov models. Ann. Statist. 30 962-1030. · Zbl 1033.60008 · doi:10.1214/aos/1031689015
[20] Roozbehani, H. and Polyanskiy, Y. (2014). Algebraic methods of classifying directed graphical models. In 2014 IEEE International Symposium on Information Theory 2027-2031.
[21] Shpitser, I., Evans, R., Richardson, T. and Robins, J. (2014). Introduction to nested Markov models. Behaviormetrika 41 3-39.
[22] Spirtes, P., Glymour, C. and Scheines, R. (2000). Causation, Prediction, and Search, 2nd ed. Adaptive Computation and Machine Learning. Cambridge, MA: MIT Press. · Zbl 0806.62001
[23] Studený, M. (2005). Probabilistic Conditional Independence Structures. Information Science and Statistics. London: Springer. · Zbl 1070.62001
[24] Sullivant, S., Talaska, K. and Draisma, J. (2010). Trek separation for Gaussian graphical models. Ann. Statist. 38 1665-1685. · Zbl 1189.62091 · doi:10.1214/09-AOS760
[25] Tian, J. and Pearl, J. (2002). A general identification condition for causal effects. In AAAI/IAAI 567-573.
[26] van Ommen, T. and Mooij, J.M. (2017). Algebraic equivalence of linear structural equation models. In Proceedings of the 33rd Annual Conference on Uncertainty in Artificial Intelligence (UAI-17).
[27] Verma, T.S. and Pearl, J. (1991). Equivalence and synthesis of causal models. In Uncertainty in Artificial Intelligence 6 255-268. Elsevier. UCLA Cognitive Systems Laboratory, Technical Report (R-150).
[28] Weihs, L., Robinson, B., Dufresne, E., Kenkel, J., Kubjas, K., II, R.M., Nguyen, N., Robeva, E. and Drton, M. (2018). Determinantal generalizations of instrumental variables. J. Causal Inference 6.
[29] Wermuth, N. (2011). Probability distributions with summary graph structure. Bernoulli 17 845-879. · Zbl 1245.62062 · doi:10.3150/10-BEJ309
[30] Wright, S. · JFM 60.1177.04 · doi:10.1214/aoms/1177732676
This reference list is based on information provided by the publisher or from digital mathematics libraries. Its items are heuristically matched to zbMATH identifiers and may contain data conversion errors. In some cases that data have been complemented/enhanced by data from zbMATH Open. This attempts to reflect the references listed in the original paper as accurately as possible without claiming completeness or a perfect matching.