Modeling the Effect of Land-Use on Person-Mile-Traveled Using Geographically-Weighted Regression

摘要

This study contributes to the literature by developing a geographically-weightedregression (GWR) model for capturing the impacts of land use on person-miles traveled anddemonstrating its benefits over simpler methods. Travel survey and land-use data from South-East Florida were used in this analysis. The empirical results reconfirm the strong impacts ofregional accessibility, land-use mixing, and connectivity on PMT. These land-use effects wereestimated to be significant after controlling for socio-economic variables. Further, the GWRmodel demonstrates that the marginal sensitivities of PMT to various land-use attributes do varyover space. This spatial variation was particularly strong in the case of the effect of regionalaccessibility. The empirical results show that allowing for flexible trends in the parameter effectsdoes improve the models and explain a greater proportion of the variance in PMT across theregion. The study also highlights the statistical superiority of the GWR model over the globalregression models.