On Gaussian conditional independence structures

Lněnička, Radim   ; Matúš, František

Kybernetika, Tome 43 (2007) no. 3, pp. 327-342

Abstract (VO)

The simultaneous occurrence of conditional independences among subvectors of a regular Gaussian vector is examined. All configurations of the conditional independences within four jointly regular Gaussian variables are found and completely characterized in terms of implications involving conditional independence statements. The statements induced by the separation in any simple graph are shown to correspond to such a configuration within a regular Gaussian vector.

Détail

Classification : 15A15, 60E05
Keywords: multivariate Gaussian distribution; positive definite matrices; determinants; principal minors; conditional independence; probabilistic representability; semigraphoids; separation graphoids; gaussoids; covariance selection models; Markov perfectness

Bibliographie

[1] Cox D., Little, J., O’Shea D.: Ideals, Varieties, and Algorithms. An Introduction to Computational Algebraic Geometry and Commutative Algebra. Second Edition. Springer-Verlag, New York 1997 | MR | Zbl

[2] Dawid A. P.: Conditional independence in statistical theory. J. Roy. Statist. Soc. Ser. B 41 (1979), 1–31 | MR | Zbl

[3] Dawid A. P.: Conditional independence for statistical operations. Ann. Statist. 8 (1980), 598–617 | MR | Zbl

[4] Dempster A. P.: Covariance selection. Biometrics 28 (1972), 157–175

[5] Frydenberg M.: Marginalization and collapsibility in graphical interaction models. Ann. Statist. 18 (1990), 790–805 | MR | Zbl

[7] Kauermann G.: On a dualization of graphical Gaussian models. Scand. J. Statist. 23 (1996), 105–116 | MR | Zbl

[8] Kurosh A.: Higher Algebra. Mir Publishers, Moscow 1975 | MR | Zbl

[9] Lauritzen S. L.: Graphical Models. (Oxford Statistical Science Series 17.) Oxford University Press, New York 1996 | MR | Zbl

[10] Levitz M., Perlman M. D., Madigan D.: Separation and completeness properties for AMP chain graph Markov models. Ann. Statist. 29 (2001), 1751–1784 | MR | Zbl

[11] Lněnička R.: On conditional independences among four Gaussian variables. In: Proc. Conditionals, Information, and Inference – WCII’04 (G. Kern-Isberner, W. Roedder, and F. Kulmann, eds.), Universität Ulm, Ulm 2004, pp. 89–101

[12] Matúš F.: Ascending and descending conditional independence relations. In: Trans. 11th Prague Conference on Information Theory, Statistical Decision Functions and Random Processes (S. Kubík and J. Á. Víšek, eds.), Vol. B, Academia, Prague, and Kluwer, Dordrecht 1991, pp. 189–200

[13] Matúš F.: On equivalence of Markov properties over undirected graphs. J. Appl. Probab. 29 (1992), 745–749 | MR | Zbl

[14] Matúš F.: Conditional independences among four random variables II. Combin. Probab. Comput. 4 (1995), 407–417 | MR | Zbl

[15] Matúš F.: Conditional independence structures examined via minors. Ann. Math. Artif. Intell. 21 (1997), 99–128 | MR | Zbl

[16] Matúš F.: Conditional independences among four random variables III: final conclusion. Combin. Probab. Comput. 8 (1999), 269–276 | MR | Zbl

[17] Matúš F.: Conditional independences in Gaussian vectors and rings of polynomials. In: Proc. Conditionals, Information, and Inference – WCII 2002 (Lecture Notes in Computer Science 3301; G. Kern-Isberner, W. Rödder, and F. Kulmann, eds.), Springer, Berlin 2005, pp. 152–161 | Zbl

[18] Matúš F.: Towards classification of semigraphoids. Discrete Math. 277 (2004), 115–145 | MR | Zbl

[19] Matúš F., Studený M.: Conditional independences among four random variables I. Combin. Probab. Comput. 4 (1995), 269–278 | Zbl

[20] Pearl J.: Probabilistic Reasoning in Intelligent Systems: Networks of Plausible Inference. Morgan Kaufmann, San Mateo, Calif. 1988 | MR | Zbl

[21] Prasolov V. V.: Problems and Theorems in Linear Algebra. (Translations of Mathematical Monographs 134.) American Mathematical Society 1996 | MR | Zbl

[22] Studený M.: Conditional independence relations have no finite complete characterization. In: Trans. 11th Prague Conference on Information Theory, Statistical Decision Functions and Random Processes (S. Kubík and J. Á. Víšek, eds.), Vol. B, Academia, Prague, and Kluwer, Dordrecht 1991, pp. 377–396

[23] Studený M.: Structural semigraphoids. Internat. J. Gen. Syst. 22 (1994), 207–217 | Zbl

[24] Studený M.: Probabilistic Conditional Independence Structures. Springer, London 2005

[26] Šimeček P.: Classes of Gaussian, discrete and binary representable independence models have no finite characterization. In: Prague Stochastics (M. Hušková and M. Janžura, eds.), Matfyzpress, Charles University, Prague 2006, pp. 622–632

[27] Šimeček P.: Gaussian representation of independence models over four random variables. In: Proc. COMPSTAT 2006 – World Conference on Computational Statistics 17 (A. Rizzi and M. Vichi, eds.), Rome 2006, pp. 1405–1412

[28] Whittaker J.: Graphical Models in Applied Multivariate Statistics. Wiley, New York 1990 | MR | Zbl