摘要: |
The National Highway Traffic Safety Administration (NHTSA) uses imputation methods to predict the values of missing alcohol data based upon characteristics of existing data. In the case of alcohol use by drivers, objective measures are available for many cases, and these known values can be used as a basis for imputing missing data. In the case of driver fatigue, objective data are not available, so other methodological and statistical approaches need to be developed to gain a greater understanding of the risks of drowsy driving and incidence of related crashes. By combining multiple methods of determining driver drowsiness and measuring related factors, we will explore the potential of utilizing methods such as those used in alcohol data imputation to generate sound estimates and cross-validate drowsy driving. A number of datasets are currently available that lend themselves to exploring crash numbers and the driving risks associated with drowsy driving. One of the most promising datasets for exploring drowsy driving is from the 2nd Strategic Highway Research Program (SHRP2) naturalistic study. By exploring and linking variables and estimates in the SHRP2 dataset and others, such as NHTSA�s crash databases, NHTSA hopes to create new statistical models and cross-validate findings in order to establish more reliable and valid estimates of drowsy driving crash risks and incidence. |