This paper considers computation of fitted values and marginal effects in the Box-Cox regression model. Two methods, 1 the “smearing” technique suggested by Duan (see Ref. [10]) and 2 direct numerical integration, are examined and compared with the “naive” method often used in econometrics.