Model transfer provides an alternative to undertaking complete data collection and model development in every planning context. The effectiveness of the transferred model in the application context can be improved by updating selected model parameters by using limited data from the application context. The effect of model updating on the transferability of disaggregate travel choice models both within and between urban areas is examined. It is found that transfer effectiveness improves with updating alternative specific constants and improves further with updating the parameter scale for both intraurban and interurban transfers. Further, the sample size necessary to obtain a substantial improvement in model transferability is a small fraction of that needed to estimate a complete model in the application context. Thus, it appears that model transfer with updating may be preferable to either full model transfer or new model estimation in situations of constrained resources.
Abstract