Modeling heterogeneity in trip-distributions with partial information

Abstract

In this paper we propose a modified gravity model that takes into account that a population generally consists of heterogeneous groups, and we suggest a new statistical test for heterogeneity. We apply our new model to two real world data sets, and it turns out that this new model fits the data surpricingly well. Not only is the effect of heterogeneity strongly significant, the model also provides far better fits than traditional trip-distribution models on these particular data sets.