“…By allowing the effects of exogenous explanatory factors to vary across individual crashes (or segments of population), more efficient, precise, and richer insights can be obtained. To account for unobserved heterogeneity, a broad spectrum of studies have successfully used different methodological alternatives including random parameter models (Anastasopoulos and Mannering, 2009;Zhao and Khattak, 2015;Alarifi et al, 2017;Bhat et al, 2017;Khattak et al, 2019;Khattak and Fontaine, 2020), correlated random parameter models (Fountas et al, 2018a;Fountas et al, 2019;Wali et al, 2019a;Matsuo et al, 2020), random parameter models with heterogeneity-in-means (Venkataraman et al, 2014;Behnood and Mannering, 2017b;Wali et al, 2018c;Hamed and Al-Eideh, 2020), random parameter models with heterogeneity-in-means and variances (Behnood and Mannering, 2017a;Seraneeprakarn et al, 2017;Xin et al, 2017;Behnood and Mannering, 2019;Al-Bdairi et al, 2020;Yu et al, 2020), latent-class models (Eluru et al, 2012;Behnood et al, 2014;Shaheed and Gkritza, 2014;Yasmin et al, 2014a;Fountas et al, 2018b), latent class models with random parameters (Xiong and Mannering, 2013), Markov-switching models Khattak and Wali, 2017), Markov-switching models with random parameters (Xiong et al, 2014), and copula based approaches (Eluru et al, 2010;Yasmin et al, 2014b;Wang et al, 2015a;Wali et al, 2018a;Wali et al, 2018e;Wang et al, 2019). For a detailed discussion on the...…”