From statistical evidence to evidence of causality