Prediction models for pedestrian and cyclist accidents.

This paper discusses a study undertaken for Transfund to produce generalized linear models for the major cycle and pedestrian accident types. A sample of traffic signal, roundabout and mid-block locations was selected in Christchurch, Hamilton and Palmerston North; three flat cities within New Zealand with a relatively high number of cyclists. The mid-block selection criteria focused on arterial routes with ‘strip’ shopping. Pedestrian and cycle count data was collected at each site, to accompany motor-vehicle count data. Data was collected on a number of non-flow variables, including visibility, number of approach lanes, crossing distance, and compliance with signalised crossing ‘green man’. Using the collected and existing data, accident prediction models were developed using generalised linear models for accidents between cyclists and motor vehicles and pedestrians and motor vehicles. These models are presented in this paper. (a) For the covering entry of this conference, please see ITRD abstract no. E212706.