A SAMPLING STRATEGY FOR REAR-END PRE-CRASH DATA COLLECTION

The involvement of a driver in a rear-end crash and the manner in which his/her vehicle collides with other vehicle(s) depends not only on the driver’s perception of the complex scenario that emerges prior to the crash, but also on his/her precrash driving behavior, response to the imminent crash situation, and performance in resolving the driving conflicts. Obviously, any effort directed toward crash countermeasures must start from data collection in order to have a better understanding of these driver-related parameters in ‘naturalistic’ settings. This would require deploying vehicles on the roadways that are equipped with certain devices to record data on such parameters. Due to the random nature of these crashes, the number of vehicles actually required to obtain the desired amount of information may be large, thereby making data collection an expensive proposition. A sample design is needed that can conserve resources and yet obtains the maximum information. The present study aims at proposing a sampling strategy for designing an optimal sample. It consists of stratifying the target population by driver’s age and allocating the sample over the strata on the basis of driver’s propensity of being in the striking/struck role in a rear-end crash. With a specific requirement of observing certain numbers of drivers in these two roles, the proposed strategy is compared with other methods of allocation, such as equal and proportional. The sample designed through this strategy is found to be much more economical in terms of the number of vehicles that need to be equipped and the number of voluntary drivers required.