IDENTIFYING A CAUSE IN MODELS WHICH ARE NOT SIMPLE BAYESIAN NETWORKS