Validating the weights in rule-based expert systems: a statistical approach