Distributional equivalence and structure learning for bow-free acyclic path diagrams (Q1689000)

Bow-free acyclic path diagrams (BAPs) are obtained from the underlying directed acyclic graphs under the condition that not both a directed and bidirect edge can be between the same pair of the considered variables. In this manuscript the first structure learning algorithm for BAPs is presented. The unknown parameters of the model are estimated by using the maximum likelihood estimation method and the RICF method. Some necessary and sufficient conditions are obtained for distributional equivalence in BAP models. The effectiveness of the introduced method is checked through some empirical results.