Ordered stated preference data including the degree of individuals' preference were interviewed and some characteristics of mode choice models based on those data were investigated to estimate the mode choice behaviour after the introduction of the New Transit System in Hiroshima. First, we compared the reliability of models based on the stated preference data (SP models) with the conventional models on the revealed preference data (RP models). Although SP models were less accurate than RP models to estimate their actual behavior, the difference of both models were not statistically significant. Secondly, in an attempt to examine the bias caused by the fatigue and inertia for three times repetitive questions, the MNL models developed by data sets from the first to the last question were successively specified. This led to a fact that the repetitive answered bias was not significant, which means that the repetitive questions were important to save our efforts to collect the data associated with stated preference. Finally, we introduced the ordered logit models which were able to incorporate the informations related to all ordered data sets. The signs of parameters of ordered logit models were logically adequate and the model efficiency of these models was shown as good as that of conventional MNL models.