Citation

Yue L, Abdel-Aty M, Lee J, Farid A. Transp. Res. Rec. 2019; 2673(2): 743-757.

Copyright

(Copyright © 2019, Transportation Research Board, National Research Council, National Academy of Sciences USA, Publisher SAGE Publishing)

DOI

10.1177/0361198119825834

PMID

unavailable

Abstract

The main objective of this study is to quantify the safety impacts of signalization at Florida's rural three-leg and four-leg stop-controlled intersections by estimating crash modification factors. The intersections are those in which stop signs are provided for the minor approaches or all-way stop-controlled intersections. The crash modification factors (CMF) are estimated using the cross-sectional method. Generalized linear models (GLM) and multivariate adaptive regression spline models (MARS) are employed with four years of Florida crash data. The K-nearest neighbor (KNN) and K-means clustering algorithms are implemented to identify the comparison sites which are sites having similar characteristics as those of the converted intersections. Furthermore, the quasi-induced exposure method is used to evaluate separately the safety effects of signalization for elderly and non-elderly drivers. According to the results, signalization contributes to an increase in property damage only (PDO) and rear-end crashes. In addition, elderly drivers are more at risk of being involved in such crashes than non-elderly drivers. In particular, at rural four-leg two-way stop-controlled intersections, signalization decreases crash severity, and a greater percentage of the decrease is observed for the elderly drivers than non-elderly especially when the intersection has a high level of major-road average annual daily traffic (AADT) and elderly driver proportion. This study also demonstrates that the MARS model shows a better model fit than the GLM model due to its strength in capturing nonlinear relationships and interaction effects among variables. This study's findings have implications for both practitioners and researchers.

Language: en