Abstract
We study the problem of restricting a Markov equivalence class of maximalancestral graphs (MAGs) to only those MAGs that contain certain edge marks,which we refer to as expert knowledge. Such a restriction of the Markovequivalence class can be uniquely represented by a restricted essentialancestral graph. Our contributions are several-fold. First, we prove certainproperties for the entire Markov equivalence class including a conjecture fromAli et al. (2009). Second, we present several new sound graphical orientationrules for adding expert knowledge to an essential ancestral graph. We also showthat some orientation rules of Zhang (2008b) are not needed for restricting theMarkov equivalence class with expert knowledge. Third, we provide an algorithmfor including this expert knowledge and show that in certain settings theoutput of our algorithm is a restricted essential ancestral graph. Finally,outside of the specified settings, we provide an algorithm for checking whethera graph is a restricted essential graph and discuss its runtime. This work canbe seen as a generalization of Meek (1995) to settings which allow for latentconfounding.