01373547

Component

The focus of this paper is to develop and assess generalized linear models based on negative binomial distribution (to account for observed over-dispersion) to estimate the number of crashes at intersections for two different scenarios. While models were developed considering all variables (including traffic volume) that are not correlated to each other as independent variables in the first scenario, models were developed considering all variables (excluding traffic volume) that are not correlated to each other as independent variables in the second scenario. Data collected for 150 intersections randomly selected in the city of Charlotte, North Carolina were used to develop the models for each scenario. The numbers of crashes at each intersection was used as a dependent variable. Demographic characteristics (population and the number of households), socio-economic characteristics (average household income and employment), and land use characteristics (commercial, industrial, institutional and residential area) within the vicinity of each intersection as well as on-network characteristics (signalized or unsignalized intersection, whether the intersection is skewed or not, the number of approaches, speed limit for major and minor street, the number of left turn, through and right turn lanes along major and minor street, and traffic volume) were considered as independent variables. Results obtained indicate that models with traffic volume have better predictive capability than those without traffic volume. An assessment of models based on the effect of buffer width indicates that a 0.25-mile buffer width around an intersection would be ideal to capture spatial off-network data and yields statistically meaningful results for both the considered scenarios. The methodology and the models could be used by practitioners to estimate potential risk at “new” intersections or at existing intersections near “new” developments so as to pro-actively apply safety treatments.

01362476

12-2880

English

14p

2012

Transportation Research Board 91st Annual Meeting

Digital/other

Figures; References; Tables

Binomial distributions; Demographics; Intersections; Land use; Traffic safety; Traffic volume

Charlotte (North Carolina)

Highways; Safety and Human Factors; I82: Accidents and Transport Infrastructure

Transportation Research Board Annual Meeting 2012 Paper #12-2880

TRIS, TRB

Feb 8 2012 5:12PM