straints that are imposed, the price vectors chosen to generate the data, and the functional form used for the cost function approximation. It is also fair to say that if input-output coefficients are the ones of interest, since these are obtained for each of the points used in the generation, it may be better to seek some approximating functions for these directly rather than seek an approximating function for the cost function and then derive the input-output coefficients. All these comments do not imply that the pseudo data approach should be given up. It is, however, important to sort out the aims of the analysis. There are some problems (like studying the effects of changes in environmental regulations) where one cannot get any answers from time series and one has to use the process analysis models that have detailed specification of technology and the constraints. But for this the appropriate thing is to do a simulation analysis of the process model itself and not seek a single equation approximation of the complex technology. What distin,guishes Griffin's approach from the garden variety simulation analyses of process models is this distillation in a single equation.