ABSTRACT
Discrete or nominal scale data often play a dominant role in transportation because many interesting policy-sensitive analyses deal with such data. Examples of discrete data in transportation include the mode of travel (automobile, bus, rail transit), the type of vehicle owned, and the type of a vehicular accident (run-off-road, rear-end, head-on, etc.). From a conceptual perspective, such data are classified as those involving a behavioral choice (choice of mode or type of vehicle to own) or those simply describing discrete outcomes of a physical event (type of vehicle accident). The methodological approach used to statistically model these is often identical, but the theory underlying these perspectives can be quite different. Discrete models of behavioral choices are derived from economic theory, often leading to additional insights in the analysis of model estimation results, whereas models of physical phenomena are derived from simple probabilistic theory. This chapter presents the derivation of discrete outcome models as well as their various estimation nuances. Examples are provided (multinomial logit and nested logit alternatives) to demonstrate the application and limitations of such models. This chapter provides the fundamentals needed to apply these models to a wide variety of transportation data analyses.
