Empirical Baye's method of identification of hazardous road location

 Empirical Baye's method of identification of hazardous road location

• An empirical method involves the use of objective, quantitative observation in a systematically controlled, replicable situation, in order to test or refine a theory.
• Empirical is something that is based completely on experiments or experience. It is the opposite of theoretical derivations which are based on theories developed in the past.Additionally, theory is constrained by assumptions and conditions
• The Empirical Baye's (EB) method increases the precision of estimates when the data is limited and corrects the regresion-to-mean bias.
• Currently, it is used in Comprehensive Highway Safety Data Model (CHSDM)
• The safety of a road is estimated by the number of accidents expected to occur in a specified period.
• If safety estimates are based on accident counts, the data may be too imprecise to be useful
• The Empirical Baye's (EB) method for estimation of safety increases the precision for estimation and corrects for regression-to-mean bias
• The EB method should be used whenever the need to estimate road safety arises
• The expected accident frequency at similar roads is determined by Safety Performance Function (SPF).
• Determination of sites that are more dangerous than others helps in better scheduling road safety policies
• Bayesian estimation for the model via a Markov Chain Monte Carlo (MCMC) approach is used
• Methods that can measure and produce comparable results concerning the risk of each site are of special interest for designing new roads or to enforce rules.
• This implies the existence of criteria that consider a particular site as hazardous
• Black-spots are dangerous locations where several accidents occur
• Treating black-spots is a well-known method of improving road safety
• From a statistical point of view, road accidents are treated as random events
• Accident count follows the Poisson Probability law. However, to correct Poisson variation in accident counts, negative binomial regression models have been used.
• Recently, Bayesian techniques have been used to tackle problems in traffic safety
• Data forms the basis for using the EB method. Hence, the previous data about hot spot accident sites on a stretch of a road are selected.
• The accident sites are classified according to their localization.
• The total number of accidents at the locations are identified
• At each site i, the number of accidents Xi is counted
• The injuries are classified into fatalities (Yi), heavily injured (WIi) and light injuries (Zi)
• Considering ti = 1 for all i=1........n the cost of accidents at a particular site (Ci) is given by       

Ci = E(Yi) + 0.075E(Wi) + 0.0035E(Zi)

E = Expenses related to a death or injury

• The problem of ranking sites is a difficult one since accidents are rare events and observed data are not necessarily a good indication as they extract data from an underlying density distribution
• Currently, the hierarchical bayesian procedure (Empirical Bayesian method) for ranking hazardous sites takes into account not only fatalities but also injuries (both light and severe) and combines this information by means of a cost function in order to rank the sites.
• It follows logically that the sites incurring the maximum cost are identified as hazardous road locations.