Citation

Flask T, Schneider IV W. Safety Sci. 2013; 53: 1-10.

Copyright

(Copyright © 2013, Elsevier Publishing)

DOI

10.1016/j.ssci.2012.08.008

PMID

unavailable

Abstract

Estimation of frequency for single-vehicle motorcycle crashes may be difficult due to challenges associated with underreporting of single-vehicle crashes and the unavailability of motorcycle-specific data. To help address these difficulties, this study focuses on modeling single-vehicle motorcycle crashes in Ohio using a full Bayesian negative binomial model with mixed effects, creating a model structure that accounts for some of the uncertainty inherent in the data. The fixed effects considered in this study incorporate geometric, administrative, and traffic information into the model while considering the limitations of motorcycle data, such as a lack of consistent, descriptive measures of motorcycle-specific traffic. The same data set is analyzed with varying levels of information describing the areas closest to each township. The Deviance Information Criterion and spatial correlation coefficient show that township level spatial random effects significantly improve the estimate of the parameters. The results show that a researcher can apply this methodology to single vehicle motorcycle crashes to find the influence of similar parameters in a given region.