Some statistical comments on the errors in NIR calibrations