Commenting some of the participating papers it is pointed out: (1) that the accuracy that can be reached in a model is limited by the number of observed trips used in the calibration, and is often exaggerated; (2) that the distance function achieved after aggregation of two trip purposes is not a mean of the two original functions but normally falling slower than any of them, and (3) that the models must be based on variables and coefficients that can be forecast.
Abstract