Blaming suicide on NASA and divorce on margarine: the hazard of using cointegration to derive inference on spurious correlation