Experience with using the Box-Cox transformation when forecasting economic time series : A comment