Abstract by Timo Pew
Justification for Considering Zero-Inflated Models in Intersection Safety Analysis
To reduce motor vehicle crashes at intersections in the state of Utah, researchers have developed statistical models for the Utah Department of Transportation (UDOT) to identify intersections experiencing significantly more crashes than expected based on roadway attributes, or ``hot spots.'' One challenge of modeling intersection related crash data in Utah is the high proportion of sites with zero crashes. There is a general reluctance to use zero-inflated models in the traffic safety literature. The primary purpose of this project is to provide a thorough explanation for why we feel that zero-inflated models can be considered a viable option for modeling crash counts. Secondarily, we will compare the goodness-of-fit and prediction abilities of implementations of zero-inflated Poisson, zero-inflated negative binomial, and negative binomial-Lindley Bayesian hierarchical models with the most recent intersection related crash dataset for the state of Utah.