Probability approach for ranking high-accident locations

Abstract

Several methods for ranking high accident locations have been considered. Although the budget constrains is the primary factor affecting the number of locations to be considered in future improvement programs, some techniques based on either accident frequency and/or accident rate are typically used. This paper presents a statistical technique employing the theoretical relationship concept of Poisson and binomial distributions. It is known that counting distributions, in particular Poisson and Binomial distributions, have been used in traffic research for a long time. Based on a certain definition for a hazardous location, the paper gives an illustration on how to rank such locations by computing a list of probabilities. The analyst can use one of these probabilities in order to select some locations to be candidates for future improvement. The paper suggests a generic format for the technique described in the study.