Severity Analysis of Crashes Using Structural Equation Modeling
Date
2020-03-02
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
ORCID
Type
Thesis
Degree Level
Masters
Abstract
Population growth, increased travel demand and, consequently, increased motor vehicle use has led to concerns about road safety in today’s society. In transportation engineering, road safety levels are measured through frequency and severity of motor vehicle crashes. Crash data has been used in road safety modeling to analyze factors that may reduce crash frequency and severity.
Regarding crash severity analysis, modeling techniques have mainly attempted to incorporate road and traffic factors into a statistical model, building a direct relationship between independent and dependent (crash severity) variables. However, some explanatory variables can affect crash severity indirectly through one or more mediating variable. Moreover, while traditional techniques have only included measured variables, there might also be unobserved factors not included in the observed data affecting crash severity. Therefore, this thesis is aimed at investigating both observed and unobserved factors that influence the severity of crashes, directly and indirectly, using a statistical technique known as structural equation modeling (SEM). Two types of crashes that affect road safety in urban and rural areas were investigated in this thesis: red-light running related (RLR) crashes and wildlife-vehicle crashes (WVC), respectively. An SEM model was developed for each crash type.
In effect, three unobserved variables were hypothesized for RLR crashes: pre-crash travel speed (TS) of the bullet vehicle (at fault), the kinetic energy (KEs) applied from the bullet vehicle to the subject vehicle(s), and crash severity. Similarly, three latent variables were introduced for WVCs: driver’s speeding attitude (SA), driver’s visibility impairment (VI), and crash severity. The results show that crash data supports the main hypothesis, with measured/latent variables adequately predicting crash severity. Regarding the RLR data, results show that both TS and KEs positively influence the overall crash severity, and that TS increase could positively affect KEs. Regarding the WVC data, the results showed that both SA and VI positively influenced overall crash severity, and that higher VI would negatively affect SA, which would indirectly decrease crash severity. Overall, these findings could help transportation practitioners to prioritize strategies and countermeasures aimed at reducing crash severity outcomes at urban and rural road sites.
Description
Keywords
Transportation Safety, Crash Severity Analysis, Structural Equation Modeling, Crash Analysis, Statistical Methods, Data and Information Technology, Safety Data Analysis and Evaluation, Highway Safety Performance, Statistical Modeling, SEM
Citation
Degree
Master of Science (M.Sc.)
Department
Civil and Geological Engineering
Program
Civil Engineering