Comparison of alternative methods for updating disaggregate logit mode choice models

Abstract

An empirical assessment of alternative methods of updating disaggregate travel choice models so that their transferability from the estimation context within which they were originally developed to an application context (which differs from the original estimation context geographically or temporally, or both) is presented. The case study for the empirical tests performed is a long-term temporal transfer of work trip logit mode choice models estimated using 1964 data for the greater Toronto area (GTA) to represent 1986 work trip mode choice in the GTA. Three updating procedures that have been previously presented in the literature are examined (Bayesian updating, transfer scaling, and combined transfer estimation), plus a fourth new procedure, joint context estimation. All four procedures assume that a small data set of observed travel choices is available for the application context, which can be used in the updating procedure. The case study results indicate that the latter three procedures all possess merit as potential updating methods, with the choice among the three depending on such items as model specification and application context sample size. The results also indicate that if the application context sample size exceeds 400 to 500 observations, then updating may provide little or no improvement over simple estimation of an application context model, especially if full model specification is supported by the available data.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 35 citations exist upstream.