Abstract in English:

Abstract: Spatial effects are intrinsic to studies on travel behavior, including modal choice problems. Among spatial dependence and heterogeneity, the resulting spatial variation of estimated coefficients is one of the most significant gains in local spatial models. Thus, this article aims to analyze a modal choice problem, considering spatial dependence and variability simultaneously. Additionally, a test is proposed to evaluate and validate the spatial variability, obtaining disaggregated results. A database adapted from the household origin-destination survey was used, which was carried out in 2007/2008 in the city of São Carlos - São Paulo, Brazil. The proposed spatial variability test uses the estimated parameters of the GWLR model (main database and 200 spatially randomized databases), compared to the confidence intervals of the coefficients in the non-spatial logit model for the spatial variability hypothesis. The results of the proposed test are similar to the reference test in the case study. The disaggregated results can be used to verify if there are certain subgroups that are more likely to be spatially stationary and if these groups exhibit a spatial pattern. Moreover, it can be observed that the local spatial model provides a better fit and estimates when compared to the non-spatial model.