Multivariate random parameters zero-inflated negative binomial regression for analyzing urban midblock crashes

摘要: Abstract Urban midblock crashes are influenced mainly by traffic operation and roadway geometric features. In this paper, 10-year crash data from 1,506 directional urban segments in Nebraska were analyzed using the multivariate random parameters zero-inflated negative binomial model to account for unobserved heterogeneity produced correlations across segments, collision types, excessive zero crashes, over dispersion. The was superior many common frequency models terms of both goodness fit prediction accuracy. Compared with fixed model, identified fewer key influencing factors revealed segment-specific effects these on different types. It showed that number lanes, annual average daily per lane, segment length might have non-positive frequencies. Segments a speed limit 45 mph had than did those lower limits, there Omaha Lincoln. also found neither presence shoulder, on-street parking, or one-way traffic, nor lane width significant influences These findings informative transportation agencies take correct efficient measures accommodate diverse demands without reducing safety. By contrast, results consistent intuition, but insufficient provide actionable recommendations.