Alternative Markov properties for chain graphs (Q2722302)

An alternative Markov property (AMP) for chain graphs (CGs) is introduced and shown to be the Markov property satisfied by a block-recursive linear system with multivariate normal errors. This study has been started by the authors [in: Uncertainty in artificial intelligence: Proceedings of the twelfth conference (eds. F. Jensen and E. Horvitz), 40-48 (1996)]. For a CG that is either an undirected graph (UG) or an acyclic directed graph (ADG), the Lauritzen-Wermuth-Frydenberg (LWF) [see \textit{S. L. Lauritzen} and \textit{N. Wermuth}, Ann. Stat. 17, No. 1, 31-57 (1989; Zbl 0669.62045); \textit{M. Frydenberg}, Scand. J. Stat. 17, No. 4, 333-353 (1990; Zbl 0713.60013)] and AMP Markov properties coincide. We note that LWF is the Markov property for adicyclic graphs, ``chain graphs'' that generalize the classical Markov properties for both UGs and ADGs. For a general CG the AMP in some ways seems to be a more direct extension of the ADG Markov property than does the LWF property. In the general case, necessary and sufficient conditions are given for the equivalence of the LWF and AMP Markov properties of a CG, for the AMP Markov equivalence of two CGs, for the AMP Markov equivalence of a CG to some ADG or decomposable UG, and for other equivalences. For CGs, in some ways the AMP property is a more direct extension of the ADG Markov property than is the LWF property.