Quality of prediction: A critique of Bronner

In this comment on Bronner's article ‘Decision style in transport mode choice’ two problems are discussed. The first and minor one deals with the choice of criterion in the model tests for the quality of prediction. The second and major problem concerns the invalidity of the model tests, which is due to the psychometrically incorrect classification procedure adopted by Bronner. This issue is extensively discussed because some of Bronner's conclusions can be shown to evolve from his incorrect assignment procedure.

As an alternative method of attitude assignment the use of a matching constant is suggested. In order to evaluate predictive quality of models, it is recommended to consider mode choice as the criterion variable instead of attitudes within travel mode categories.