On calculating the edgeworth approximate distribution of an econometric estimator or test statistic