Application of a novel hybrid multigroup statistical approach to investigate the factors affecting crash severity

Abstract

Identifying the complex relationships contributing to crash severity is vital for effective road safety strategies but can be challenging. This study explores a hybrid Structural Equation Modeling/Fuzzy-set Qualitative Comparative Analysis (SEM-FsQCA) technique to analyze these relationships, including moderation effects. By integrating SEM and FsQCA to offer a more comprehensive analysis, it overcomes a key challenge of traditional methods—the inability to simultaneously address complex causal relationships and interaction effects. Also investigated was the potential of the Synthesizing Minority Oversampling Technique (SMOTE) for addressing the inherently imbalanced nature of the crash severity and other data used for the analysis. Utilizing a database of Ohio collector roads as a case study, a multigroup analysis was also implemented to analyze factors in lower and higher-income neighbourhoods, which were characterized by imbalanced samples, and assess how combinations of road and environmental variables affect crash severity on roads adjacent to these two neighbourhoods. The SEM results indicated that, regardless of the neighbourhood income level, age, percentage of grade, the proportion of the population having a diploma or higher, horizontal curve, and speed limit all significantly affect crash severity. Those results did indicate that the effects of independent and moderating variables are significantly different for the two neighbourhoods. Using FsQCA, the causal configurations leading to higher crash severity were explored for the two neighbourhood categories. The results of the case study revealed that crash prevention measures could be more effectively developed for crashes based on the income level of neighbourhoods adjacent to the collector roads investigated.