Investigation the Effect of Pavement Condition Characteristics on Bend Segments Accident Frequency: Application of Fixed and Random Parameters Negative Binomial Models

摘要

Fixed and random parameter negative binomial models were employed in this paper to analyze the impact of pavement condition properties on accidents at two bends site categories (based on radius of curvature). Six-years (2005-2010) of accident data were considered at bends site categories on road network of Norfolk County in the U.K. The first objective is to examine the performance of the fixed and random parameter negative binomial model approaches. The second objective is to investigate the effects of skid resistance combined with traffic characteristics (annual average daily flow, heavy vehicles percentage, and speed limit) and geometric characteristics (radius of curvature, gradient, number of lanes, and number of minor access), on accident numbers. The results suggest that there was a significant improvement in log-likelihood when using random parameters models. Regarding the effects of the explanatory variables, the results reveal that the annual average daily flow, percentage of heavy vehicles, gradients and number of minor access, and rut depth are positively associated with more accidents. Speed limit, radius of curvature, number of lanes, skid resistance, and texture depth are found to be significant variables decreasing the numbers of accidents.