On the Equivalence of Causal Models