If one time-series model is used for forecasting a low demand item and another time-series model is used for forecasting a high demand item, which measure of forecast accuracy provides a valid comparison of the errors from these two time-series?