added a research item
Accident Prediction Models
Accident Prediction Models have been developed for urban and rural intersections and links. This includes models for traffic signals, roundabout, priority and uncontrolled intersections, rural highways, motorways, urban arterials , collector and local streets. Accident (1995 to 1999) and traffic data at over 1000 sites throughout New Zealand were used to develop the models. Test statistics have been prepared showing the goodness-of-fit and confidence intervals of the models. Application of the models in economic appraisal (evaluation) has been discussed along with the changes that would need to be made to the (NZ) Project Evaluation Manual to incorporate the models.
Accident prediction models have been developed for urban and rural intersections and links and for other sites, including isolated curves, narrow bridges, single lane bridges and railway crossings. This paper details the generalised linear models that have been developed from reported injury accident data and traffic counts in New Zealand for major accident types and total accidents. Refinements have been made to the methods used to calculate the goodness-of-fit testing statistic, the deviance function to address the low mean value problem. Applications of the models are discussed, including economic evaluation, performance measures and safety management systems, and optimisation networks flow patterns to improve safety.