Multilevel Dirichlet process mixture analysis of railway grade crossing crash data

Benchmarking regions using a heteroskedastic grouped random parameters model with heterogeneity in mean and variance: Applications to grade crossing safety analysis

“…A number of traffic safety studies accounted for group effects (instead of site effects) in random effects, random parameters, and latent class modeling to capture variations in unknown/unmeasured factors that vary across groups (Wu et al, 2013;Heydari et al, 2014a;Heydari et al, 2016b;Sarwar et al, 2017;Fountas et al, 2018a and2018b;Cai et al, 2018). A discussion on the importance of accounting for group (e.g., region) effects in traffic safety research is provided in Heydari et al, 2016b. For example, the latter study discusses how spatially and non-spatially related unobserved factors can be captured to some extent through accounting for group (e.g., region) specific effects.…”

“…With regard to highway-railway grade crossing safety issues, several studies discuss various aspects of crossing safety (Saccomanno et al, 2004;Millegan et al, 2009;Chaudhary et al, 2011;Chadwick et al 2014;Tung et al, 2015;Wang et al, 2016a;Heydari et al, 2016b;Haleem, 2016;Liu and Khattak, 2017;Hsu and Jones, 2017;Sperry et al, 2017;Zhang et al, 2017;Larue et al, 2018;Beanland et al, 2018). Many grade crossing studies can be divided into two categories: (1) crash-frequency studies (Hauer and Persaud, 1987;Austin and Carson, 2002;Saccomanno and Lai, 2005;Park and Saccomanno, 2005;Oh et al, 2006;Yan et al, 2010;Medina and Benekohal, 2015;Heydari and Fu, 2015;Lu and Tolliver, 2016;Heydari et al, 2016a;Heydari et al, 2017b; Guadamuz-Flores and Aguero-Valverde, 2017); and (2) crash-consequence studies (Eluru et al, 2012;Hao et al, 2015;Ghomi et al, 2016;Zhao et al, 2018).…”

“…While a macro-level approach can provide insights that are useful for planning and policy development (Washington et al, 2010), it may suffer from some shortcomings relating to the aggregation of data, which could, for instance, lead to ecological fallacy (Davis 2004). Previous work on grade crossing safety with a focus on making macro-level inferences is rare (Truong et al, 2016;Heydari et al, 2016b). To our knowledge, only Heydari et al (2016b) draw macro-level inferences for grade crossings using disaggregates crossing-level crash data.…”

“…Note that Heydari et al, (2016b), using a Dirichlet process mixture approach, found that the normality assumption in the random effects does not hold for Canadian grade crossings equipped with flashing lights and bells. Here, we did not find any evidence of non-normality, perhaps because in this paper we are analyzing all Canadian public grade crossings.…”

A flexible discrete density random parameters model for count data: Embracing unobserved heterogeneity in highway safety analysis

“…In fact, various efforts have been underway to mitigate the unobserved heterogeneity problem in traffic safety analysis, which could be summarized as follows: (1) the random effects (intercepts) approach (Shankar et al, 1998;Kim al., 2007;Aguero-Valverde, 2013;Naznin et al, 2016;Sarwar et al, 2017b); (2) the random parameters (also referred to as random slopes or random coefficients) approach (Anastasopoulos and Mannering, 2009;El-Basyouny and Sayed, 2009;Venkataraman et al, 2014;Wu et al, 2013;Anastasopoulos, 2016;Sarwar et al, 2017a;Alarifi et al, 2017;Bogue et al, 2017;Chen et al, 2017;Bhat et al, 2017, Fountas andAnastasopoulos, 2017;Shaon et al, 2018;Fountas et al, 2018;Cai et al, 2018;Heydari et al, 2018); (3) the finite mixture approach (Park and Lord, 2009;Park et al, 2016;Yasmin and Eluru, 2016;Zou et al, 2017); (4) the finite mixture random parameters approach (Xiong and Mannering, 2013;Li et al, 2018); and recently (5) the Bayesian semiparametric Dirichlet process approach (Heydari et al, 2016a and2016b;Shirazi et al, 2016;Yu et al, 2016;Heydari et al, 2017;Cheng et al, 2018), which has been applied only in random effects model settings in traffic safety research. A detailed discussion relating to various statistical models used in traffic safety research and unobserved heterogeneity can be found in Lord and Mannering (2010), Mannering and Bhat (2014), and…”