Analytic Methods in Accident Research最新文献

This study explores how heterogeneity has been examined in transportation safety analyses, specifically focusing on latent class modeling, which has gained popularity and has successfully captured unobserved heterogeneity. The study firstly identifies a large volume of relevant papers in the safety analysis domain and analyzes how models have been used by focusing on key elements of the latent class model (along with the proposed typology of segmentation-based heterogeneity models). In the literature, various class-specific outcome models have been used. They are determined by the type of outcome variable and are also highly associated with the analysis context. For example, crash severity and crash likelihood/frequency analyses are the main applications where crash severity is often treated as binary, nominal, or ordered, whereas crash likelihood/frequency is subject to count data or survival data modeling. The study reviews the number of classes selected in empirical applications and how they were determined. It is found that in safety analyses, it is more common to choose the number of classes based on the judgement of the analyst than quantitative measures (e.g., BIC). This implies that we value interpretability of the latent class model and solutions with many classes (i.e., greater model complexity, many parameters) often hinder the interpretation of models. This paper also covers further discussions about heterogeneity including model comparisons (homogeneity models versus latent class models and random parameters versus latent class models), modeling intra-class heterogeneity, possible alternative model specifications that have been rarely used in the literature, and issues related to temporal instability.

Injury severity in police crash reports is usually recorded in three to five classes, including property damage, slight, moderate, serious, and fatal injuries. Among these classifications, serious injuries are commonly classified as cases where a road user is admitted to a hospital. In this classification, the length of hospital stay is not differentiated, whether one day or ten days, as long as the road user has been admitted to the hospital. As such, the inferences drawn from assuming that all serious injuries (1 if a road user is admitted to the hospital; 0 otherwise) are at the same severity level inherently suffer from aggregation bias and may not provide a thorough understanding of this severity category. This study proposes a hazard-based duration modelling approach to examine the severity of serious injury crashes measured in a continuous spectrum. Specifically, using the length of hospital stay as the measure of serious injuries, a random parameters hazard-based duration model with heterogeneity in means was applied to model serious injury crashes obtained by linking crash records in police and hospital databases. To address temporal instability, the injury records sources from Abu Dhabi, United Arab Emirates (UAE), between 2015 and 2019 were modelled separately for each year. The results showed that factors positively associated with more serious injury severity (prolonged length of hospital stay) are rural areas, high posted speed limits of 100–160 km/h, overturned crashes, speeding, impaired driving, involvements of a heavy vehicle, nighttime crashes, lack of restraint usage, and injuries to the head or lower extremities. In particular, speeding violations during nighttime are positively associated with more serious injuries. Furthermore, the means of the random parameters of head injury are positively influenced by speeding, lack of restraint usage, and motorcycle involvement through the heterogeneity-in-mean specification of the hazard-based duration model. The proposed modelling approach to model serious traffic injuries using a hazard-based duration model provides a comprehensive understanding of the factors associated with serious injuries.

Variable speed limits are frequently used to improve traffic safety and harmonise traffic flow. This study investigates how, and to what extent, drivers reduce their speed upon passing a variable speed limit sign. We specifically consider the impact on braking behaviour due to the systematic inclusion of different social pressures exerted by surrounding traffic. This social pressure is the natural result of having two vehicle cohorts created by a change in the variable speed limit (the new speed limit being higher than the original). The cohort with the higher speed limit overtakes vehicles with the lower speed limit, instigating a specific passing rate on drivers in the lower speed cohort. A driving simulator study is employed to obtain individual driver data whilst being able to systematically change the social pressure applied. A sample comprising sixty-seven participants conducted multiple randomised drives, with varying passing rates from as low as 90 veh/h to as high as 360 veh/h. The speed reduction behaviour of the participants is modelled using a random parameter duration modelling approach. Both the panel nature of the data and unobserved heterogeneity are captured through a correlated grouped random parameters with heterogeneity-in-the-mean model. The random parameters are predicated on the different passing rate scenarios, allowing drivers to take shorter or longer to reduce their speeds compared to the reference passing rate. It is shown that the extent of social pressure impacts braking behaviour and therefore affects safety measures, which is a function of the magnitude of the speed limit change. In addition, an extensive decision tree analysis is conducted to understand differential braking behaviour. Results reveal that, on average, female drivers take a shorter time to reduce their speed under a high passing rate but longer in a low passing rate scenario compared to males. Similarly, young drivers are found to take longer to reduce their speeds in a high passing rate scenario compared to other age groups. Our main findings indicate that the within-cohort safety is lowest under low passing rates due to comparatively larger speed differences between drivers. Yet, under a high passing rate, we observe an increase in violation of the speed limit by the lower speed limit vehicles (but less within cohort speed differences). Whilst normally this would be an undesired effect across cohorts, this violation is argued to lead to increased safety due to the smaller discrepancy in speed.

Driving at nighttime may make drivers more likely to be involved in fatal crashes. To investigate the temporal instability and age differences of contributors determining different injury severity levels in nighttime crashes, this paper estimates three groups of random parameters logit models with heterogeneity in the means and variances (young/middle-age/old groups). Nighttime single-vehicle crashes in this study are gathered over four years in California, from January 1, 2014, to December 31, 2017, provided by Highway Safety Information System, including single-vehicle crashes occurring under dark, dawn, and dusk lighting conditions. Simultaneously, to investigate the temporal instability and transferability of nighttime crash severity relating to drivers of different ages, three disaggregate groups are defined: young drivers (15–29 years old), middle-age drivers (30–49 years old), old drivers (over 49 years old). Three injury-severity categories are determined as outcome variables: severe injury, minor injury, and no injury, while multiple factors are investigated as explanatory variables, including driver characteristics, vehicle characteristics, roadway characteristics, environmental characteristics, crash characteristics, and temporal characteristics. Two series of likelihood ratio tests are undertaken to unveil the contributors determining nighttime crash injury severities varying among drivers of different ages over time. Besides, the current study also compares the differences between out-of-sample and within-sample predictions. The results indicate the unstable direction of predictions across different age groups over time and underscore the necessity to adequately accommodate the temporal instability and age differences in accident prediction. More studies can be conducted to accommodate the self-selectivity issue and the out-of-sample prediction differences between using the parametric models and non-parametric models.

This study aims to address the questions of how driving risk evolves during car-following processes and what factors contribute to the underlying evolution patterns. An empirical study is conducted using real world car-following data collected in the Shanghai Naturalistic Driving Study (SH-NDS). The evolution of the driving risk induced by the dynamic coupling between the leading and following vehicles during the car-following process is characterized by how an instantaneous crash-risk measure - rear crash risk index (RCRI) - changes by time. A spectral clustering analysis is first conducted to classify the driving risk evolution of the observed car-following maneuvers, showing the existence of five distinctive risk evolution patterns in the car-following processes. In order to investigate the relationship between the identified driving risk evolution clusters and their contributing factors, a regression analysis employing a random parameter multinomial logit model with heterogeneity in means and variances is followed, revealing several significant contributing factors to the car-following risk evolution patterns, such as congestion level, driver’s ability to maintain stable headways, and vehicle deceleration. This study has provided important insights into driving risk from the new perspective of risk evolution patterns, which is expected to have significant implications for the future development of advanced traffic management and traveler information systems (ATMS/ATIS) strategies, advanced driver assistance systems (ADAS), and connected and autonomous vehicles (CAV).

To minimize non-recurrent congestion, a better understanding of the factors that affect accident clearance time is crucial, in order to optimize incident management strategies. A number of methods have been developed to predict incident clearance duration, but few of those have considered the time-varying nature of certain observed factors. In addressing this gap in the literature, this study developed a grouped random parameters hazard-based duration model with time-varying covariates, while accounting for unobserved heterogeneity. Data on accidents, traffic, road inventory, and real-time weather condition were compiled for the Kaiyang freeway in 2014. Comparison of candidate models shows that the proposed model with Weibull distribution exhibits the best fit performance. The results suggest that the effects of rear-end accident, involvements of trucks or other vehicles, evening hours, and shoulder blockage on the hazard function are heterogeneous across observations. Other variables such as angle accident, injury severity, traffic volume and composition, morning or pre-dawn hours, and blockage of overtaking lane were also found to have significant but homogenous effects on accident clearance time. More importantly, the results also reveal the significant effects of the time-varying covariates (wind speed, temperature, and humidity). Accordingly, the viability and superiority of the proposed model in analyzing accident clearance time are confirmed. Overall, the results of this study are expected not only to improve traffic incident management by allowing government agencies to better understand factors affecting accident clearance times, but also to facilitate incident clearance through the recognition of time-varying pattern.

Research in highway safety continues to struggle to address two potentially important issues; the role that unobserved factors may play on resulting crash and injury-severity likelihoods, and the issue of identification in safety modeling caused by the self-selective sampling inherent in commonly used safety data (the fact that drivers in observed crashes are not a random sample of the driving population, with riskier drivers being over-represented in crash data bases). This paper addresses unobserved heterogeneity using mixing distributions and attempts to provide insight into the potential sample-selection problem by considering data before and during the COVID-19 pandemic. Based on a survey of vehicle usage (vehicle miles traveled) and subsequent statistical modeling, there is evidence that riskier drivers likely made up a larger proportion of vehicle miles traveled during the pandemic than before, suggesting that the increase in injury severities observed during COVID-19 could potentially be due to the over-representation of riskier drivers in observed crash data. However, by exploring Florida crash data before and during the pandemic (and focusing on crashes where risky behaviors were observed), the empirical analysis of observed crash data suggests (using random parameters multinomial logit models of driver-injury severities with heterogeneity in means and variances) that the observed increase in injury severity during the COVID-19 pandemic (calendar year 2020) was likely due largely to fundamental changes in driver behavior and less to changes in the sample selectivity of observed crash data. The findings of this paper provide some initial guidance to future work that can begin to more rigorously explore and assess the role of selectivity and resulting identification issues that may be present when using observed crash data.

Pedestrians represent a vulnerable road user group at signalised intersections. As such, properly estimating pedestrian crash risk at discrete short intervals is important for real-time safety management. This study proposes a novel real-time vehicle-pedestrian crash risk modelling framework for signalised intersections. At the core of this framework, a Bayesian Generalised Extreme Value modelling approach is employed to estimate crash risk in real-time from traffic conflicts captured by post encroachment time. A Block Maxima sampling approach, corresponding to a Generalised Extreme Value distribution, is used to identify pedestrian conflicts at the traffic signal cycle level. Several signal-level covariates are used to capture the time-varying heterogeneity of traffic extremes, and the crash risk of different signal cycles is also addressed within the Bayesian framework. The proposed framework is operationalised using a total of 144 hours of traffic movement video data from three signalised intersections in Queensland, Australia. To obtain signal cycle-level covariates, an automated covariate extraction algorithm is used that fuses three data sources (trajectory database from the video feed, traffic conflict database, and signal timing database) to obtain various covariates to explain time-varying crash risk across different cycles. Results show that the model provides a reasonable estimate of historical crash records at the study sites. Utilising the fitted generalised extreme value distribution, the proposed model provides real-time crash estimates at a signal cycle level and can differentiate between safe and risky signal cycles. The real-time crash risk model also helps understand the differential crash risk of pedestrians at a signalised intersection across different periods of the day. The findings of this study demonstrate the potential for the proposed real-time framework in estimating the vehicle-pedestrian crash risk at the signal cycle level, allowing proactive safety management and the development of real-time risk mitigation strategies for pedestrians.

A copula-based dependence approach accommodates various facets of dependence structures in building multivariate stochastic models. In existing studies, applications of copula for ordinal random variables are predominantly modeled by employing traditional ordered models (ordered logit/probit) while assuming the effects of parameters to remain the same across all observations. The methodological contributions of this study are grounded in addressing the abovementioned significant methodological gaps in the application of copula formulation by proposing a copula-based random parameters nominal-ordinal joint model construct of correlated random variables. Specifically, we propose and develop a random parameters binary logit-generalized ordered logit copula formulation while also complementing the proposed approach by accommodating the effects of unobserved heterogeneity in parameter estimates. To the best of the authors’ knowledge, this study is the first instance to incorporate generalized ordered formulation within copula in extant econometrics literature. Further, to obtain a direct effect of exogenous variables on dependence, we parameterize the copula dependence structure as a function of different covariates in six different copula structures including a wide range of dependency structures which represent radial symmetry and asymmetry, and asymptotic tail dependence. The empirical contributions of this study are grounded in the application of the proposed copula-based formulation by examining ‘active traveler (pedestrian and bicyclist) crash type’ and ‘active traveler injury severity outcomes’ as two dimensions of active travel injury severity mechanism. The model is estimated by using crash data for the years 2012 through 2018 from the state of Queensland, Australia, by employing a comprehensive set of exogenous variables. In addition, the analyses are further augmented by complementing the elasticity effects of exogenous variables.

In conventional safety analysis, traffic and crash data are often aggregated at the geographical units like census tracts, street blocks, and traffic analysis zones, which are often delineated by roads and other physical entities. A considerable proportion of crashes may occur at or near the boundary of geographical units. Such the crashes, also known as boundary crashes, can correlate with the explanatory variables of neighboring geographical units, regardless of the spatial proximity. This could then bias the parameter estimation of crash frequency model. In this study, a novel data-driven approach is developed for the allocation of boundary crashes. For example, crash severity and bicyclist characteristics are considered in the crash feature-based allocation. An illustrative case study based on built environment, population, traffic and bicycle crash data from 289 Lower Layer Super Output Areas (LSOAs) of London in the period 2017–2019 was conducted. Results indicate that high matching percentages of boundary crash allocation can be achieved. Furthermore, prediction performances, in terms of root mean square error (RMSE) and mean absolute error (MAE), of the crash frequency models based on the proposed crash feature-based allocation method is superior, compared to that based on conventional boundary crash allocation methods like half-and-half and iterative assignment approaches. Last but not least, more influencing factors that affect the bicycle crash frequency at macroscopic level can be identified. Findings should be indicative to the spatial safety analysis for different geographical configurations.