e branch-and-bound technique. Several settings for thresholds are investigated.
Since the consideration set is not observable, it can be argued that the parameters of the consideration set construction model and the choice model shall be estimated together. This is done by Kaplan and Prato (2010). The consideration set Cn for trav- eler n is derived from the master set using a conjunctive heuristic semi-compensatory model: the probability to find consideration set Cn is given by the probability than respondent n uses a specific set of thresholds for the independent values. If an in- dependent variable is out of range w.r.t. a threshold, the corresponding route is not considered. The thresholds are unknown in advance and they are estimated using a maximum likelihood method. The authors only consider the route length and the number of turns.