Traditional approaches to modeling injury severity (multinomial logit model, nested logit model) presume that the effect of each parameter is fixed across all crashes (Kim et al.,2013; Wu et al., 2014), but fail to capture the influences of unobserved heterogeneity (Jalayer et al., 2018). The random parameters model allows the model parameters to vary in crashes, thereby capturing the heterogeneity of influencing factors. (Behnood and Mannering, 2017a, Behnood and Mannering, 2017b,