Abstract by Christian Davis
Hot Spot Identification Analysis Using a Fully Bayesian Spatial Generalized Linear Mixed Model
In recent years, the Utah Department of Transportation (UDOT) has not only aimed to diminish, but eradicate fatalities caused by motorized vehicle crashes. As a result, many studies have been conducted to understand the relationship of important road and driving characteristics with the likelihood of a crash. However, incorporating spatial random effects using a fully Bayesian framework remains an underdeveloped aspect of modeling crash data. In this project, we use a fully Bayesian spatial generalized linear mixed model to understand and facilitate roadway safety in three ways. First, we strive to more accurately understand the effect of roadway and driving features on the number of crashes by accounting for the heterogeneity across Utah roads. Second, we incorporate a spatial random effect to improve predictions and better quantify the uncertainty associated with the effects of road construction. Finally, we use the estimates of the spatial random effects to identify road segments that seem to be more dangerous than would be expected by only accounting for the effects of road and driving characteristics.