Discrete choice models with latent explanatory variables using subjective data

Traditional discrete choice models for travel behavior use only manifest explanatory variables such as travel time and cost. Choice decisions, however, are strongly affected by latent factors such as comfort, safety, and reliability. Some authors have reported that using subjective ratings against those latent factors as explanatory variables gives better fit to the calibration data, but such an approach is rarely useful in forecasting. This paper proposes a methodology for incorporating choice data and subjective rating data to develop integrated models of choice behavior and psychological perception structure. The proposed models are based on the multiple-indicator-multiple-cause type model and the LISREL type model.