Use of Structural Equation Modeling to Measure Severity of Single-Vehicle Crashes

Abstract

Injury severity and vehicle damage are two of the main indicators of the level of crash severity. Other factors, such as driver characteristics, roadway conditions, highway geometry, environmental factors, vehicle type, and roadside objects, may also be directly or indirectly related to crash severity. All these factors interact in such complicated ways that it is often difficult to identify their interrelationships. The aim of this study was to examine the relationships between these contributors and the severity of single-vehicle crashes. Structural equation modeling (SEM) offers the opportunity to explore the complex relationships between variables by handling endogenous variables and exogenous variables simultaneously. Furthermore, SEM allows latent variables to be included in the model and bridges the gap between dependent and explanatory variables. In this study, the number of latent variables was defined by the understanding of collision force, kinetic energy, and mechanical process of a collision, as well as statistical goodness of fit that was based on available data. Three SEM models (one with one latent variable, one with two, and one with three) representing the hypothesized relationships between collision force, speed of a vehicle, and severity of a crash were developed and evaluated in an attempt to unravel the relationships between exogenous factors and severity of single-vehicle crashes. On the basis of goodness of fit and model predictive power, the model with two latent variables outperformed the other two. Additional insights about model selection were provided through the development and comparison of the three models.