Accident Models for Two-Lane Rural Segments and Intersections

Abstract

Data collected from the states of Minnesota and Washington on rural two-lane highways are used to build accident models for segments and three-legged and four-legged intersections stop-controlled on the minor legs. The quantity, quality, and variety of data collected, together with the advanced techniques applied in the analysis, make this study of special interest. Variables include traffic, horizontal and vertical alignments, lane and shoulder widths, roadside hazard rating, channelization, and number of driveways. Models are of negative binomial and extended negative binomial form and yield R2 values from 0.42 to 0.73 and overdispersion parameters from 0.20 to 0.51. A segment model combining both states and including state as a variable, and intersection models derived from Minnesota data, are featured, along with summary statistics, goodness-of-fit measures, and cross-validation between the states. Segment accidents depend significantly on most of the roadway variables collected, while intersection accidents depend primarily on traffic. The study recommends development of adjustment factors for different regions and times and further development of extended negative binomial models.

On the nature of over-dispersion in motor vehicle crash prediction models

“…While for intersection crash models almost all studies considered the traffic volume from major and minor road, other studies took the effect of geometric factors in addition to the exposure variables to develop the mean structure. For exam ple, Bauer and Harwood (2000), Vogt and Bared (1998), , Lyon et al (2003) considered geometric factors in addition to traffic volume. In all of these studies they adopted a logarithmic transformation of the major road and minor road traffic flow along with a linear effect of common geometric fac tors.…”

A spatially disaggregate analysis of road casualties in England

“…Poisson distributions assume the variance is equal to the mean, a condition that is frequently violated and is known as overdispersion. Negative binomial models provide are a generalization of the Poission model that can account for this (Miaou, 1994;Shankar et al, 1995;Vogt and Bared, 1998). Although the source of overdispersion in count data cannot be distinguished, its presence can be adjusted by introducing a stochastic component in the log-linear relationship between the expected numbers of accident in an observation unit i, i μ and the covariates X…”

Estimation of Safety Effectiveness of Composite Shoulders on Rural Two-Lane Highways

“…In order to estimate the expected crashes in the treatment group had no treatment been applied, the EB estimated crash frequency in the before period in a treatment group is multiplied by the ratio. 19 …”

“…Many model forms have been recorded by researchers (Miaou (16), Vogt and Bared (19)) and the negative binomial specification has become the forerunner in crash count modeling. For this specific research, negative binomial regression models were applied both in the cross-sectional approach and to develop SPFs in the EB approach.…”

“…The base SPFs in the HSM were based on a study by Vogt and Bared (19). They collected roadway geometric features for rural two-lane highways in Washington and Minnesota.…”