抄録
The present study is a part of continued development of discrete choice models for location analysis. The discrete choice modeling paradigm, and in particular the logit model, are research topics that have been continuously developed and refined for years in the field of transportation applications. Modeling locational choices, on the other hand, differs from modeling transportation choices in that geographically referenced data is used, whereby the choices are specifically spatial. The utility function of the proposed model is specified with autoregressive expressions for the deterministic and error components, and the model is evaluated with reference to three alternative models: the standard logit model, a logit model with an autoregressive deterministic term, and the mixed logit model with autoregressive error terms. The methodological development was reported in the previous paper. The present study investigates the applicability of those models in terms of zone and area unit system and this paper reports the results. It is found that the relative distance between zones is the major effect to determine the applicability of the proposed model. In the other word, the model shows an improved performance over the three reference models when the choices exhibit relative difference in distance between zones.
