zbMATH — the first resource for mathematics

Chain graph models and their causal interpretations (with discussion). (English) Zbl 1090.62103
Summary: Chain graphs are a natural generalization of directed acyclic graphs and undirected graphs. However, the apparent simplicity of chain graphs belies the subtlety of the conditional independence hypotheses that they represent. There are many simple and apparently plausible, but ultimately fallacious, interpretations of chain graphs that are often invoked, implicitly or explicitly. These interpretations also lead to flawed methods for applying background knowledge to model selection. We present a valid interpretation by showing how the distribution corresponding to a chain graph may be generated from the equilibrium distributions of dynamic models with feed-back. These dynamic interpretations lead to a simple theory of intervention, extending the theory developed for directed acyclic graphs. Finally, we contrast chain graph models under this interpretation with simultaneous equation models which have traditionally been used to model feed-back in econometrics.

MSC:
 62M99 Inference from stochastic processes 05C90 Applications of graph theory 05C20 Directed graphs (digraphs), tournaments
Software:
 [1] Andersson, S. A. , Madigan, D. and Perlman, M. D. (1996) An alternative Markov property for chain graphs. In Proc. 12th Conf. Uncertainty in Artificial Intelligence (eds F. V. Jensen and E. Horvitz), pp. 40â48. San Francisco: Morgan Kaufmann. [2] ââ(1997) A characterization of Markov equivalence classes for acyclic digraphs. Ann. Statist., 25, 505â541. · Zbl 0876.60095 [3] ââ(2001) Alternative Markov properties for chain graphs. Scand. J. Statist., 28, 33â85. · Zbl 0972.60067 [4] Arnold, B. , Castillo, E. and Sarabia, J. M. (1999) Conditionally Specified Distributions. New York: Springer. · Zbl 0749.62002 [5] Bentzel, On recursiveness and interdependency in economic models, Rev. Econ. Stud. 22 pp 153– (1954) [6] Besag, Spatial interaction and the statistical analysis of lattice systems (with discussion), J. R. Statist. Soc. B 36 pp 302– (1974) · Zbl 0327.60067 [7] ââ(1974b) On spatial-temporal models and Markov fields. In Trans. 7th Prague Conf. Information Theory, Statistical Decision Functions and Random Processes, pp. 47â55. Prague: Academia. [8] ââ(1975) Statistical analysis of non-lattice data. Statistician, 24, 179â195. [9] Bollen, K. A. (1989) Structural Equations with Latent Variables. New York: Wiley. · Zbl 0731.62159 [10] Box, Use and abuse of regression, Technometrics 8 pp 625– (1966) · Zbl 0203.21503 [11] Cooper, G. F. (1995) Causal discovery from data in the presence of selection bias. In Preliminary Pap. 5th Int. Wrkshp AI and Statistics, Jan. 4thâ7th, Fort Lauderdale (ed. D. Fisher), pp. 140â150. [12] Cowell, R. G. , Dawid, A. P. , Lauritzen, S. L. and Spiegelhalter, D. J. (1999) Probabilistic Networks and Expert Systems. New York: Springer. · Zbl 0937.68121 [13] Cox, Design of experiments and regression, J. R. Statist. Soc. A 147 pp 306– (1984) · Zbl 0547.62048 [14] Cox, Linear dependencies represented by chain graphs (with discussion), Statist. Sci. 8 pp 204– (1993) [15] ââ(1996) Multivariate Dependencies: Models, Analysis and Interpretation. London: Chapman and Hall. · Zbl 0880.62124 [16] ââ(2000) On the generation of the chordless four-cycle. Biometrika, 87, 204â212. · Zbl 0976.62054 [17] Darroch, Markov fields and log-linear interaction models for contingency tables, Ann. Statist. 8 pp 522– (1980) · Zbl 0444.62064 [18] Dawid, Conditional independence in statistical theory (with discussion), J. R. Statist. Soc. B 41 pp 1– (1979) · Zbl 0408.62004 [19] ââ(2000) Causal inference without counterfactuals. J. Am. Statist. Ass., 95, 407â448. · Zbl 0999.62003 [20] Edwards, The analysis of contingency tables by graphical models, Biometrika 70 pp 553– (1983) [21] Fisher, A correspondence principle for simultaneous equation models, Econometrica 38 pp 73– (1970) [22] Frydenberg, The chain graph Markov property, Scand. J. Statist. 17 pp 333– (1990) · Zbl 0713.60013 [23] Geman, Stochastic relaxation, Gibbs distributions, and the Bayesian restoration of images, IEEE Trans. Pattn Anal. Mach. Intell. 6 pp 721– (1984) · Zbl 0573.62030 [24] Gibbs, W. (1902) Elementary Principles of Statistical Mechanics. NewHaven: Yale University Press. · JFM 33.0708.01 [25] Gilks, W. R. , Richardson, S. and Spiegelhalter, D. J. (1996) Markov Chain Monte Carlo in Practice. New York: Chapman and Hall. · Zbl 0832.00018 [26] Goldberger, Structural equation models in the social sciences, Econometrica 40 pp 979– (1972) [27] Grenander, Representations of knowledge in complex systems (with discussion), J. R. Statist. Soc. B 56 pp 549– (1994) · Zbl 0814.62009 [28] Haavelmo, The statistical implications of a system of simultaneous equations, Econometrica 11 pp 1– (1943) · Zbl 0063.01836 [29] Hammersley, J. and Clifford, P. (1971) Markov fields on finite graphs and lattices. Unpublished. [30] Hastings, Monte Carlo sampling methods using Markov chains and their applications, Biometrika 57 pp 97– (1970) · Zbl 0219.65008 [31] Heckerman, Dependency networks for inference, collaborative filtering, and data visualization, J. Mach. Learn. Res. 1 pp 49– (2000) [32] Hofmann, R. (2000) Inference in Markov blanket networks. Technical Report FKI-235-00. Technical University of Munich, Munich. [33] Hofmann, R. and Tresp, V. (1998) Non-linear Markov networks for continuous variables. In Advances in Neural Information Processing Systems 10 (eds M. I. Jordan, M. J. Kearns and S. A. Solla), pp. 521â527. Cambridge: MIT Press. [34] Jensen, F. V. (1996) An Introduction to Bayesian Networks. London: University College London Press. [35] Kiiveri, H. and Speed, T. P. (1982) Structural analysis of multivariate data: a review. In Sociological Methodology (ed. S. Leinhardt). San Francisco: Jossey-Bass. [36] Kiiveri, Recursive causal models, J. Aust. Math. Soc. A 36 pp 30– (1984) [37] Koster, Markov properties of non-recursive causal models, Ann. Statist. 24 pp 2148– (1996) · Zbl 0867.62056 [38] ââ(1999) Linear structural equations and graphical models. Lecture Notes. Fields Institute, Toronto. [39] ââ(2000) Marginalizing and conditioning in graphical models. Technical Report. Erasmus University, Rotterdam. [40] Lauritzen, S. L. (1996) Graphical Models. Oxford: Clarendon. · Zbl 0907.62001 [41] ââ(1999) Generating mixed hierarchical interaction models by selection. Technical Report R-99-2021. Department of Mathematical Sciences, University of Aalborg, Aalborg. [42] ââ(2001) Causal inference from graphical models. In Complex Stochastic Systems (eds O. E. Barndorff-Nielsen, D.âR. Cox and C. KlÃ$$\tfrac14$$ppelberg), pp. 63â107. Boca Raton: Chapman and HallâCRC. · Zbl 1010.62004 [43] Lauritzen, Local computations with probabilities on graphical structures and their application to expert systems (with discussion), J. R. Statist. Soc. B 50 pp 157– (1988) · Zbl 0684.68106 [44] Lauritzen, S. L. and Wermuth, N. (1984) Mixed interaction models. Technical Report R 84-8. Institute for Electronic Systems, Aalborg University, Aalborg. · Zbl 0669.62045 [45] ââ(1989) Graphical models for associations between variables, some of which are qualitative and some quantitative. Ann. Statist., 17, 31â57. · Zbl 0669.62045 [46] MatÃ$${}^o$$Å!, Conditional independences among four random variables III: Final conclusion, Combin. Probab. Comput. 8 pp 269– (1999) · Zbl 0941.60004 [47] Meek, C. (1995) Causal inference and causal explanation with background knowledge. In Proc. 11th Conf. Uncertainty in Artificial Intelligence (eds P. Besnard and S. Hanks), pp. 403â410. San Francisco: Morgan Kaufmann. [48] Metropolis, Equations of state calculations by fast computing machines, J. Chem. Phys. 21 pp 1087– (1953) [49] Mohamed, The determinants of infant mortality in Malaysia: a graphical chain modelling approach, J. R. Statist. Soc. A 161 pp 349– (1998) [50] Neyman, J. (1923) On the Application of Probability Theory to Agricultural Experiments: Essay on Principles. (in Polish) (Engl. transl. D. Dabrowska and T. P. Speed, Statist. Sci., 5 (1990), 465â480). [51] Ord, K. (1976) An alternative approach to modelling linear systems. Unpublished. [52] Pearl, J. (1988) Probabilistic Inference in Intelligent Systems. San Mateo: Morgan Kaufmann. · Zbl 0746.68089 [53] ââ(1993) Graphical models, causality and intervention. Statist. Sci., 8, 266â269. [54] ââ(1995) Causal diagrams for empirical research. Biometrika, 82, 669â710. · Zbl 0860.62045 [55] ââ(1998) Graphs, causality, and structural equation models. Sociol. Meth. Res., 27, 226â284. [56] ââ(2000) Causality: Models, Reasoning, and Inference. Cambridge: Cambridge University Press. · Zbl 0959.68116 [57] Preston, Generalised Gibbs states and Markov random fields, Adv. Appl. Probab. 5 pp 242– (1973) · Zbl 0318.60085 [58] Richardson, T. S. (1996) Models of feedback: interpretation and discovery. PhD Thesis. Carnegie-Mellon University, Pittsburgh. [59] ââ(1998) Chain graphs and symmetric associations. In Learning in Graphical Models (ed. M.âJordan), pp. 231â260. Dordrecht: Kluwer. · Zbl 0926.62001 [60] ââ(2001) Chain graphs which are maximal ancestral graphs are recursive causal graphs. Technical Report 387. Department of Statistics, University of Washington, Seattle. [61] Richardson, T. S. and Spirtes, P. (2000) Ancestral graph Markov models. Technical Report 375. Department of Statistics, University of Washington, Seattle. · Zbl 1033.60008 [62] Ripley, B. (1981) Spatial Statistics. New York: Wiley. · Zbl 0583.62087 [63] Roberts, Exponential convergence of Langevin distributions and their discrete approximation, Bernoulli 2 pp 341– (1996) · Zbl 0870.60027 [64] Robins, A new approach to causal inference in mortality studies with sustained exposure periodsâ application to control of the healthy worker survivor effect, Math. Modllng 7 pp 1393– (1986) · Zbl 0614.62136 [65] Rubin, Estimating causal effects of treatments in randomized and non-randomized studies, J. Educ. Psychol. 66 pp 688– (1974) [66] Speed, A note on nearest-neighbour Gibbs and Markov distributions over graphs, Sankhya A 41 pp 184– (1979) · Zbl 0475.60034 [67] Spiegelhalter, Bayesian analysis in expert systems (with discussion), Statist. Sci. 8 pp 219– (1993) [68] Spirtes, P. (1995) Directed cyclic graphical representations of feedback models. In Proc. 11th Conf. Uncertainty in Artificial Intelligence (eds P. Besnard and S. Hanks), pp.â491â498. San Francisco: Morgan Kaufmann. [69] Spirtes, P. , Glymour, C. and Scheines, R. (1993) Causation, Prediction and Search. New York: Springer. · Zbl 0806.62001 [70] Spirtes, P. , Meek, C. and Richardson, T. S. (1995) Causal inference in the presence of latent variables and selection bias. In Proc. 11th Conf. Uncertainty in Artificial Intelligence (eds P.âBesnard and S. Hanks), pp. 403â410. San Francisco: Morgan Kaufmann. [71] Spirtes, P. and Richardson, T. S. (1997) A polynomial-time algorithm for determining DAG equivalence in the presence of latent variables and selection bias. In Preliminary Pap. 6th Int. Wrkshp AI and Statistics, Jan. 4thâ7th, Fort Lauderdale (eds D. Madigan and P. Smyth), pp. 489â501. [72] Spirtes, Using path diagrams as a structural equation modeling tool, Sociol. Meth. Res. 27 pp 182– (1998) [73] Spitzer, F. (1971) Random Fields and Interacting Particle Systems. Washington DC: Mathematical Association of America. · Zbl 0241.60094 [74] Strotz, Recursive versus nonrecursive systems: an attempt at synthesis, Econometrica 28 pp 417– (1960) [75] StudenÃ$$\tfrac12$$, On chain graph models for description of independence structures, Ann. Statist. 26 pp 1434– (1998) · Zbl 0930.62066 [76] Verma, T. and Pearl, J. (1990) Equivalence and synthesis of causal models. In Proc. 6th Conf. Uncertainty in Artificial Intelligence (eds P. Bonissone, M. Henrion, L. N. Kanal and J. F. Lemmer), pp.â255â270. Amsterdam: North-Holland. [77] Wermuth, Block-recursive regression equations (with discussion), Rev. Bras. Probab. Estatist. 6 pp 1– (1992) [78] Wermuth, N. , Cox, D. and Pearl, J. (1994) Explanations for multivariate structures derived from univariate recursive regressions. Technical Report 94-1. University of Mainz, Mainz. [79] ââ(1999) Explanations for multivariate structures derived from univariate recursive regressions. Technical Report. University of Mainz, Mainz. [80] Wermuth, On substantive research hypotheses, conditional independence graphs and graphical chain models (with discussion), J. R. Statist. Soc. B 52 pp 21– (1990) · Zbl 0749.62004 [81] Whittaker, J. (1990) Graphical Models in Applied Multivariate Statistics. Chichester: Wiley. · Zbl 0732.62056 [82] Wold, H. O. A. (1953) Demand Analysis. New York: Wiley. · Zbl 0050.36802 [83] ââ(1954) Causality and econometrics. Econometrica, 22, 162â177. [84] Wright, Correlation and causation, J. Agric. Res. 20 pp 557– (1921) [85] Anderson, The relation between the strength and diameter of wool fibres, J. Text. Inst. 41 pp T481– (1950) [86] Andersson, S. A. , Madigan, D. and Perlman, M. D. (1996) An alternative Markov property for chain graphs. In Proc. 12th Conf. Uncertainty in Artificial Intelligence (eds F. Jensen and E. Horvitz), pp. 40â48. San Francisco: Morgan Kaufmann. [87] â(1997) A characterization of Markov equivalence classes for acyclic digraphs. Ann. Statist., 25, 505â541. · Zbl 0876.60095 [88] â(2001) Alternative Markov properties for chain graphs. Scand. J. Statist., 28, 33â85. · Zbl 0972.60067 [89] Andersson, Symmetry and lattice conditional independence in a multivariate normal distribution, Ann. Statist. 26 pp 525– (1998) · Zbl 0943.62047 [90] Angrist, J. D. and Krueger, A. B. (2000) Empirical strategies in labor economics. In Handbook of Labor Economics, vol. III, ch. 23. New York: Elsevier. [91] Besag, J. (2002) Likelihood analysis of binary data in space and time. In Highly Structured Stochastic Systems (eds P. J. Green, N. L. Hjort and S. Richardson). Oxford: Oxford University Press. To be published. [92] Cowell, R. G. , Dawid, A. P. , Lauritzen, S. and Spiegelhalter, D. J. (1999) Probabilistic Networks and Expert Systems. New York: Springer. · Zbl 0937.68121 [93] Cox, Design of experiments and regression (with discussion), J. R. Statist. Soc. 147 pp 306– (1984) · Zbl 0547.62048 [94] Cox, D. R. and Reid, N. (2000) The Theory of the Design of Experiments. New York: CRC Press. · Zbl 1009.62061 [95] Cox, Linear dependencies represented by chain graphs (with discussion), Statist. Sci. 8 pp 204– (1993) [96] â(1996) Multivariate Dependencies: Models, Analysis, and Interpretation. London: Chapman and Hall. · Zbl 0880.62124 [97] Dawid, A. P. (2002a) Influence diagrams for causal modelling and inference. Int. Statist. Rev., to be published. · Zbl 1215.62002 [98] â(2002b) Causal inference using influence diagrams: the problems of partial compliance (with discussion). In Highly Structured Stochastic Systems (eds A. Frigessi and S. Richardson). Oxford: Oxford University Press. To be published. [99] Dempster, Covariance selection, Biometrics 28 pp 157– (1972) [100] Fader, Modeling consumer choice among SKU’s, J. Marktng Res. 33 pp 442– (1996) [101] Fisher, R. A. (1935) The Design of Experiments. Edinburgh: Oliver and Boyd. [102] Friedman, M. (1968) Theory and analysis of consumption and savings functions. In Economic Statistics and Econ- ometrics (ed. A. Zellner), pp. 236â253. Boston: Little, Brown. [103] Frydenberg, The chain graph Markov property, Scand. J. Statist. 17 pp 333– (1990) · Zbl 0713.60013 [104] Gross, Do liquidity constraints and interest rates matter for consumer behavior?: evidence from credit card data, Q. J. Econ. 117 pp 149– (2002) · Zbl 1015.91519 [105] Hobert, The effect of improper priors on Gibbs sampling in hierarchical linear models, J. Am. Statist. Ass. 91 pp 1461– (1996) · Zbl 0882.62020 [106] Horvath, A discordant-sibship test for disequilibrium and linkage: no need for parental data, Am. J. Hum. Genet. 63 pp 1886– (1998) [107] Koster, Markov properties of non-recursive causal models, Ann. Statist. 24 pp 2148– (1996) · Zbl 0867.62056 [108] â(2000) Marginalizing and conditioning in graphical models. Technical Report EUR/FSW-Soc/2000.02. Erasmus University, Rotterdam. [109] Laird, The sib transmission/disequilibrium test is a Mantel-Haenszel test, Am. J. Hum. Genet. 63 pp 1915– (1998) [110] Lauritzen, S. L. (1996) Graphical Models. Oxford: Clarendon. · Zbl 0907.62001 [111] â(2000) Causal inference from graphical models. In Complex Stochastic Systems (eds O. E. Barndorff-Nielsen, D. R. Cox and C. KlÃ$$\tfrac14$$ppelberg), ch. 2, pp. 63â107. Boca Raton: CRC Press. · Zbl 1010.62004 [112] Lauritzen, Graphical models for associations between variables, some of which are qualitative and some quantitative, Ann. Statist. 17 pp 31– (1989) · Zbl 0669.62045 [113] Lindley, Parametric empirical Bayes inference: comment, J. Am. Statist. Ass. 78 pp 61– (1983) [114] Madsen, Invariant normal models with recursive graphical Markov structure, Ann. Statist. 28 pp 1150– (2000) · Zbl 1105.62346 [115] Mantel, Chi-square tests with one degree of freedom: extensions of the MantelâHaenszel procedure, J. Am. Statist. Ass. 58 pp 690– (1963) · Zbl 0114.11601 [116] Meyer, Natural and quasi-experiments in economics, J. Bus. Econ. Statist. 13 pp 151– (1995) [117] Paz, Representation of irrelevance relations by annotated graphs, Fund. Inform. 42 pp 149– (2000) · Zbl 0956.68116 [118] Pearl, J. (2000) Causality. Cambridge, Cambridge University Press. [119] Piantadosi, S. (1997) Clinical Trials: a Methodologic Perspective. New York: Wiley. · Zbl 1081.62102 [120] Risch, The future of genetic studies of complex human diseases, Science 273 pp 1516– (1996) [121] Rosenbaum, Choice as alternative to control in observational studies (with discussion), Statist. Sci. 14 pp 259– (1999) [122] â(2002) Observational Studies, 2nd edn. New York: Springer. · Zbl 0985.62091 [123] Rosenberg, M. (1968) The Logic of Survey Research. New York. Basic Books. [124] Rosenzweig, Natural ânatural experimentsâ in economics, J. Econ. Lit. 38 pp 827– (2000) · doi:10.1257/jel.38.4.827 [125] Shadish, W. R. , Cook, T. D. and Campbell, D. T. (2002) Experimental and Quasi-experimental Designs for Generalized Causal Inference. Boston: Houghton-Mifflin. [126] Spielman, A sibship test for linkage in the presence of association: the sib transmission/disequilibrium test, Am. J. Hum. Genet. 62 pp 450– (1998) [127] Spirtes, P. , Glymour, C. and Scheines, R. (1993) Causation, Prediction and Search. New York: Springer. · Zbl 0806.62001 [128] StudenÃ$$\tfrac12$$, M. (1998) Bayesian networks from the point of view of chain graphs. In Proc. 14th Conf. Uncertainty in Artificial Intelligence (eds G. F. Cooper and S. Moral), pp. 496â503. San Francisco: Morgan Kaufmann. [129] Whittaker, J. (1990) Graphical Models in Applied Multivariate Statistics. Chichester: Wiley. · Zbl 0732.62056