Underlying Rates of Binomial Distributed Traffic Accident Data
journal of kerbala university,
2007, Volume 3, Issue 2, Pages 209-216
AbstractTraffic accident data is considered as an efficient tool to identify the degree of hazard at different locations of highway system. An accurate estimation of underlying true traffic accident rate may lead to efficient and economic safety improvement program. Accident data can be considered as random variables that have Poisson or non-Poisson distributions. A regular variation of accident data may reveal to the appropriateness of the Binomial distribution. A procedure to estimate underlying true accident rate as well as optimum time-period of accident counts for Poisson process is available while it is not for non-Poisson process.This paper proposes a new procedure to estimate the upper & lower limits of underlying accident rates depending on the observed accident rate of accident data having Binomial distribution according to different confidence degrees. The procedure includes testing data for randomness and the appropriate probability distribution that fits the data.The optimum time-period of traffic accident data provides a relatively precise estimation of underlying rate of accidents and minimizes cost of data collection as well as the social-economic losses associated in traffic accidents. A time-period beyond five years shows a relatively small decrease in the proportional uncertainty of the estimated underlying rates. Hence, a time-period of five years is sufficient for the purpose of estimation in case of binomial distributed traffic accident data. The developed procedure is a statistically reliable for purposes of programs identification of hazardous locations that may depend on the true underlying rates rather than the observed rates.
- Article View: 155
- PDF Download: 133