Temporal instability assessment of motorcyclist-injury severities using categorical principal component analysis and random parameter approach with heterogeneity in means

AbstractAbstractMotorcyclists are considered as one of the most vulnerable road participants that often suffer higher injury severities. Furthermore, contributing factors of motorcyclist-injury severities may vary over time, which requires further investigation. In this study, with Michigan crash data from 2015 to 2018, categorical principal component analysis (CatPCA) is firstly conducted to assess the similarities/differences among yearly samples. Then, a random parameter logit model with heterogeneity in means is employed for each analysis year. Marginal effects are also estimated to quantify the temporal instability of the influencing factors. The results reveal that some determinants of motorcyclist-injury severities are temporally unstable across the studied years, such as middle-aged motorcyclist, helmet worn, signal control, clear weather, two-vehicle crashes, and disabling damage. However, some factors have relatively stable effects on motorcyclist-injury severities in most of the year periods, such as alcohol impaired, totally or partially ejected from the motorcycle, stopped on the roadway, and posted speed limits higher of 50 mph. The findings can help decision makers to propose cost-effective motorcycle safety improvements and policies.Keywords: motorcyclist-injury severitycategorical principal component analysisrandom parameter logit modelheterogeneity in meanstemporal instability Additional informationFundingThe study is funded by National Natural Science Foundation of China (NSFC-72271207). Special thanks to Eric F. Jiang (Vandegrift High School) for polishing the overall language of the paper.