Análise exploratória para a modelagem da frequência de acidentes de trânsito agregados ao nível de zonas de tráfego

Abstract

This study presents an exploratory analysis of the variables that may contribute to modeling of road safety at the strategic and tactical planning level. At first, a geographic database of 126 traffic zones is compiled from available socioec-onomic datasets, roadway data and geocoded accidents. Aiming to correctly partial out the effects of each variable on accident frequency, the analysis comprises: i) a phenomenological and correlation analysis that yielded a reduced set of predictor variables, and ii) a Negative Binomial regression analysis that identified the significant predictors. Given the spatial nature of geographic data, the assumption of spatial independence between model errors is analyzed. The results confirmed that accidents are related to income and economic activity, at strategic level, and physical and operation attributes of the roadway system, at the tactical level. The evidence of spatial dependence at the strategic level points to directions for further work.