Citation

Zhang C, Ivan JN, Ravishanker N. Transp. Res. Rec. 2008; 2083: 162-169.

Copyright

(Copyright © 2008, Transportation Research Board, National Research Council, National Academy of Sciences USA, Publisher SAGE Publishing)

DOI

10.3141/2083-19

PMID

unavailable

Abstract

This paper presents an investigation of a new exposure definition for same-direction collisions on two-lane rural roads. This new exposure definition, vehicle time spent in following other vehicles, is computed on the basis of the concept of percent time spent following as provided in Highway Capacity Manual (2000). In assessing the validity of this exposure measurement in crash prediction models, collisions are categorized by contributing factors. Two distinct categories emerge for same-direction collisions—(a) rear-end and (b) same-direction sideswipe and turning—from different distributions of contributing factors for observed collisions of each of these categories. Generalized linear models are used to estimate crash prediction models for each category separately and together, and different distribution assumptions used in the models are investigated to address overdispersion. According to the model results, the new exposure definition has a linear relationship with total same-direction collisions, especially rear-end collisions alone. The primary contributing factor reported for rear-end collisions is following too closely. Therefore, finding vehicle time spent following to be linearly related to these crashes suggests a key for identifying the appropriate exposure definition for any type of collision—to define the exposure so that it relates to opportunities for the corresponding contributing factor to come into play. This finding further suggests that, for development of prediction models, instead of categorizing collisions by the manner in which vehicles collide, it might be more effective to categorize them strictly by contributing factor.