The work develops methodology for score-based structure learning of directed maximal ancestral graphs as an theoretically appealing model class generalising Bayesian networks for dealing with ...