Type of Document Thesis Author Mishra, Abhishek Author's Email Address firstname.lastname@example.org URN etd-11152006-223835 Title Analysis of Run-Off-the-Road Crashes Involving Overcorrection Degree Master of Science Department Civil and Environmental Engineering, Department of Advisory Committee
Advisor Name Title Lisa K. Spainhour Committee Chair John Sobanjo Committee Co-Chair W. Vigil Ping Committee Member Keywords
- Logistic Regression
Date of Defense 2006-08-02 Availability unrestricted AbstractABSTRACT
Overcorrection crashes have long been a concern in crash analysis because the outcomes vary from the vehicle coming to rest after a rollover, to the vehicle passing through the opposite lanes either hitting or without hitting an oncoming vehicle, to rear end and sideswipe impacts. These crashes occur when a vehicle begins drifting off the road one way and the driver oversteers in the opposite direction resulting in a vehicle- vehicle, fixed object, or overturning crash. While analyzing fatal crash data from the year 2000 in the state of Florida, it was found that these crashes represented around 25 % of the total number of run-off-the-road (ROR) crashes. With that intention, this research developed models to identify the factors that influence ROR crashes involving overcorrection.
Since the outcome of these models is discrete in nature, logistic regression was identified as the most suitable approach. A set of binary logistic regression models were applied to fatal crash data, which were collected from different sources, primarily traffic homicide reports, for crashes occurring on state roads in Florida. Various results were derived from this study and the conclusions were made accordingly. It was found that female drivers are nearly 1.5 times more likely to overcorrect than males. A straight moving vehicle is less likely to be involved in an overcorrection crash than a vehicle that is moving in any other pattern. It was also found that the presence of rumble strips has a negative impact on overcorrection crashes. The odds ratio and p-value for this variable are 1.731 and 0.053 respectively. The odds ratio indicates that the presence of rumble strip increases the chances of overcorrection. All of these factors were significant at minimum at the 85 percent level. For the overcorrection crashes, an additional statistical model was developed to analyze the factors that led the vehicle to leave the paved shoulder prior to overcorrection. These factors are found to be driver sex, age, vehicle movement, vehicle type, speeding, annual daily traffic and presence of rumble strips. Based of the results of this study, recommendations were made suggesting various countermeasures, both behavioral and roadway related, to minimize the occurrence of overcorrection crashes.
Filename Size Approximate Download Time (Hours:Minutes:Seconds)
28.8 Modem 56K Modem ISDN (64 Kb) ISDN (128 Kb) Higher-speed Access Abhi_thesis.pdf 1.20 Mb 00:05:32 00:02:50 00:02:29 00:01:14 00:00:06