Question

Why should causation not be inferred from comparative or correlational designs?