This paper aims to identify properties of attrition involving Kofu shopper panels and also to develop a shopping destination model with this panel attrition. First, three logit models are calibrated; the two are for one-wave panel data set and the one is for the second progress of two-wave panel data set. Second, two types of destination choice models are estimated and compared with each other: The one concerns the weighted method considering such attrition behavior. The other is the model using non-weighted method. The results indicated that one-wave panel data set is characterized by its random property and that more sensitive parameters of this choice model can be estimated by the weighted method.
Journal of Japan Society of Civil Engineers, Ser. D3 (Infrastructure Planning and Management)