Construction of minimal d-separators in a dependency system

Rules for efficiently finding minimal d -separating sets in DAG dependency models are inferred based on the d -separation criterion and acyclic property of a digraph. The rules allow one to acceleratedly identify the presence or absence of edges in a model. These rules use only unconditional independencies and conditional independencies of the first rank. Versions of the causal faithfulness assumption are formulated that justify the use of these rules in inferring the skeleton of a model from statistical data.