Combined estimation of activity generation models incorporating unobserved small trips using probe person data

Abstract

Probe Person (PP) survey with Global Positioning Systems (GPS) and web diary enables the collection of much higher resolution data in real time. The new method of travel behavior expected to be fused with Questionnaire type travel survey, which has been the basis of travel demand data in transportation planning. In the conventional questionnaire based survey, the problem of missing trips and activities is pointed out that caused non-response bias. This study considers the mechanism of missing activities in Person Trip (PT) surveys and proposes a combined estimation method using a sample selection model. The method contributes to correcting the bias of missing data by data fusion of PP data and PT data. The empirical results show that several personal attributes and contexts of the activities are found to be significant factors of missing of activities in paper-based PT surveys.