Trip Generation Models Using Cumulative Logistic Regression

Abstract

This paper evaluates the usefulness of the cumulative logistic regression model for estimating trip generation. The cumulative logistic regression model is a type of discrete choice model that estimates relationships between an ordered dependent variable, for example, person trip generation, and a set of independent variables, for example, household size, income, and workers. In addition to testing the model form, life cycle, area type, and accessibility variables are evaluated along with a set of widely used explanatory variables. A secondary focus of this paper is on the issue of temporal stability. Temporal stability is concerned with how models developed during one period of time transfer to a future period. The evaluation includes models based on widely used explanatory variables in addition to models supplemented with life cycle, area type, and accessibility variables to evaluate whether these variables result in improved stability. Analysis includes models estimated using 1995 survey data, applied using 2006 socioeconomic data, and evaluated against 2006 observed data. The results of this analysis show that cumulative logistic regression models are good candidate models for estimating trip generation and for improving the temporal stability of the model results. With respect to life cycle, area type, and accessibility, this research shows that there is benefit in including these variables to help explain trip making and to improve temporal stability.