Chapter 7: Longitudinal Methods

A wide range of methods fall in the category of longitudinal models. For example, aggregate time-series analysis is a longitudinal method since the data used are longitudinal. More prevalent in the context of transport modeling, however, are analyses that focus on both cross-sectional variations and longitudinal changes in the phenomenon of interest. The discussion here is, focused on panel analysis, i.e., analysis that is based on combined cross-sectional and longitudinal data.

Suppose the behavior of interest can be expressed as a stochastic process, i.e., a process comprising random events that take place over time. An example is a household's vehicle holding behavior; the number and types of vehicles available to a household vary over time as the household acquires a new vehicle, replaces a vehicle it has had with a new one, or disposes of a vehicle. The behavior of interest may be represented by a set of possibly infinite discrete states, such as the type of vehicle transaction, or by a continuous variable, such as the cumulative amount of money spent for the maintenance of a household vehicle since its acquisition. In either case, it is assumed in this chapter that changes in behavior occur at discrete time points either as transitions between discrete states, or as instantaneous increases or decreases in a continuous variable value.