Abstract:
In modeling trip-chaining behavior a number of issues arise the most fundamental being the underlying decision process. This issue is critical because it affects the modeling effort directly. In this paper we argue that trip-chaining behavior is based neither on a simultaneous nor on a sequential decision process, but on a rather in-between particular type of sequential decision making process called a relatively-sequential process. Moreover, single-stop trip chains seen in the light of relatively-sequential decision processes comprise the building blocks of complex trip chains. The relatively-sequential process is further formally characterized in the tradition of the probabilstic threshold theory.
The threshold theory offers an explicit behavioral interpretation of looking at activity sequencing as an r-sequential spatial interaction process. The modeling effort deals with individual attitudes toward trip chaining yielding macro frequencies, representable by general gravity models. These frequencies are shown to be filtered frequencies of more traditional gravity models for unchained trip flows. A family of threshold gravity models is proposed corresponding to individual commuters' attitudes toward separation. In this manner, this alternative modeling framework may serve to bridge the gap between macro and micro approaches in trip-chain models since it can use data typically gathered in transportation studies.
Maximum likelihood estimation of separation-threshold gravity models based on a home interview survey show that the proposed theoretical framework has initiated a legitimate and promising research direction. Indeed, estimated trip-chain frequencies fit very well to data. Given that estimation results are based on only few observations, the very good fit of the models is an indication that threshold gravity models may not be "data-hungry". This, in turn, give another evidence that threshold gravity models although aggregate in principle may not need more data than disaggregate models to study the distribution of trip-chain frequencies. An additional very significant empirical finding is that in the presence of the data available home-to-work and the reverse single-stop trip chains may be indistinguishable in the sense that there may be no need for separate parameter estimates. This is a very pleasant result from the transportation modeler's perspective. Finally, short-term forecasting
suggestions are made under reasonable assumptions and analytic relationships for forecast trip-chain frequencies are proposed along with a method to compute them.
