“…The model assigns non-negative probabilities to all alternatives in the universal choice set including those not in an individual's true choice set. As a result, it introduces biases in parameter estimation of the utility function, which leads to inaccurate predictions (Arentze and Timmermans, 2005;Kwan and Hong, 1998;Landau et al, 1982;Pellegrini et al, 1997;Thill, 1992;Thill andHorowitz, 1997a, 1997b;Williams and Ortúzar, 1982) and thus erroneous interpretations of human behavior (Thill, 1992;Thill and Horowitz, 1997a). Pellegrini et al (1997), for example, adopted a baseline of 14 stores, and compared it with choice sets from 3 to 13 stores.…”