摘要: |
A common type of crash that occurs on the roadways is the rear-end crash. Every year a large number of drivers are involved in such crashes. In order to develop effective crash countermeasures, it is important to have a better understanding of the driving behavior and performance of a driver prior to a rear-end crash. For that purpose, experiments need to be conducted in which the drivers can be observed in 'naturalistic' settings and data can be collected on the driver-related parameters. This study discusses some of the sampling issues involved in the process of data collection in the above context. Contingency analysis is conducted to suggest criteria for stratifying the target population. A probabilistic approach is used for allocating the sample over the strata thus formed. An estimate of the number of vehicles needed to observe a specific number of rear-end crashes is obtained. This estimation problem is treated as the 'discrete waiting-time' problem. Additionally, Binomial probability distribution is used to estimate the number of drivers who would be involved in rear-end crashes as a result of deploying a certain number of vehicles. The approach adopted in this study is fairly general and can be used to resolve the sampling issues in similar setups. Two databases, the General Estimates System (GES) and the Fatality Analysis Reporting System (FARS), compiled by the National Highway Traffic Safety Agency (NHTSA), have been used in this study. |