The relevancy of the composition of the choice set is investigated by Prato and Bekhor (2007). The authors investigate the effect of the choice set generation tech- nique on the choice model parameter estimates and on the generated predictions. The study creates two choice sets (one generated by the branch-and-bound technique and one that merges results from labeling, link elimination, link penalty and stochastic link trait adaptations). Six choice models are estimated using each choice set. All models use ten explanatory variables. The effect of combining choice set generation techniques and choice models is investigated. The branch-and-bound generator uses fixed settings for the criteria used. The effect of the criteria is not investigated, proba- bly because the branch-and-bound technique already outperforms the other ones with respect to coverage.