Hybrid dimension reduction and logit models for glare-induced crash severity

Abstract Sun glare during sunrise and sunset and headlight glare at night can temporarily reduce contrast sensitivity and hazard detection, creating high risk driving conditions that contribute to severe crashes on specific roadway and lighting contexts. This study develops a two-stage, cluster-driven discrete choice framework to examine unobserved heterogeneity in glare-induced crash severity using Texas police-reported crash data from 2017 to 2024. Cluster Correspondence Analysis (CCA) is first applied to identify three glare-related crash typologies, after which cluster-specific multinomial and random-parameter logit models with heterogeneity in means are estimated to assess injury severity outcomes. The results reveal pronounced cluster-dependent severity mechanisms. Low-speed urban angle crashes are associated with substantially lower fatal and incapacitating injury risk, particularly under daylight conditions. In contrast, high-speed straight-line and rear-end crashes, especially those involving speed-control failures, exhibit the highest likelihood of severe outcomes. Rural nighttime crashes on unlit two-lane roads show elevated injury severity likelihood due to headlight dazzle, with vehicle type and lighting conditions moderating these effects. Lower speed limits and daylight consistently reduce fatal injury risk, while higher speeds, inadequate lighting, and straight-line movements are associated with higher severity. These findings demonstrate that integrating clustering with random-parameter logit modeling improves interpretability and supports targeted countermeasures such as dynamic speed management, glare-reducing lighting treatments, adaptive headlamp technologies, and focused driver education.