ABSTRACT
This study is designed to provide a new perspective for the multiday traveling mode choice analysis. A GPS-enabled activity-travel survey conducted in Shanghai, China is used to obtain the multiday traveling dataset. Complete travel records over consecutive days are available for each respondent, which are recognized as a hierarchical and nested structure, i.e. trips nested within a day and days nested within individual traveler. A multilevel and mixed-effects logistics regression model is employed, investigating not only the effect of trip-, day-, and individual-level variables on mode choice, but whether or not particular effect varies across days and individuals. Results regarding the fixed-effect part suggest that only limited individual-level predictors play a significant role, and the effect of day-of-week variable directly demonstrates the nonexistence of a ‘typical’ traveling day. More importantly, a substantial amount of random effects is revealed, which explicitly justifies the multilevel approach.
Acknowledgments
The authors would like to express their appreciation to anyone who has provided suggestions and comments on this paper.
Disclosure statement
No potential conflict of interest was reported by the author.