|
Abstract:
|
Systemic safety improvements to roadways, when selected and targeted appropriately, provide a tremendous opportunity to proactively reduce automobile crashes. As such, detailed information is needed on the crashes and contributing factors that are best targeted by systemic improvements and the types of locations and situations where these crashes occur. In this talk, we demonstrate how to model the number of car crashes along road segments using Poisson regression with spatially correlated random effects. These spatial random effects serve to account for important covariates left out of the model or potential residual spatial correlation in the crash counts. Estimating both the coefficients associated with covariates and estimated spatial components, we are be able to determine (1) which road characteristics lead to higher accident rates and (2) which road segments have higher than expected crash rates for further investigation.
|
