On Computing Probabilistic Abductive Explanations