Content uploaded by Narayan Venkataraman
Author content
All content in this area was uploaded by Narayan Venkataraman on May 14, 2018
Content may be subject to copyright.
A preview of the PDF is not available
Occupant injury severities in hybrid-vehicle involved crashes: A random parameters approach with heterogeneity in means and variances
Abstract and Figures
Differences in hybrid and non-hybrid vehicle design, and potential differences in driver-related behavior among owners of these vehicle types, can potentially have interesting implications for safety-related policies. To study possible differences in hybrid and non-hybrid occupant injury severities in motor vehicle crashes, this paper uses a sample of hybrid-vehicle-involved crashes and estimates a mixed logit model of the resulting injury level of the most severely injured occupant in the crash, while accounting for possible heterogeneity in the means and variances of model parameters. A total of 2015 crashes in Washington State, involving at least one hybrid vehicle in the 5-year period from January 1, 2006 to December 31, 2010 were analyzed. The data included crash information regarding occupants, vehicles, environmental conditions at the time of the crash, hybrid and non-hybrid vehicle attributes, crash-contributing circumstances for both hybrid and non-hybrid involved vehicles, collision type and crash location information relating to intersections, functional class of the highway, and highway curvature. Model estimation results show that a wide range of variables influence the most severely injured occupant, and that the number-of-occupants parameter and the intersection-location indicator parameter are random with significant heterogeneity in both means and variances. Sources of heterogeneity include the ratio of hybrid to non-hybrid vehicle counts in the crash, vehicle weight to horsepower ratio range (maximum difference in ratio) for the crash, number of adult occupants aged 41–64 years, functional class, and vehicle type interactions. The results further demonstrate the potential of models that address unobserved heterogeneity to unravel important relationships in the analysis of highway injury severities.
Figures - uploaded by Narayan Venkataraman
Author content
All figure content in this area was uploaded by Narayan Venkataraman
Content may be subject to copyright.
... Also, Huang et al. pointed out that the statistical analysis results could support strong heterogeneity effects in crash data (12). Based on this conclusion, Seraneeprakarn et al. further validated the influence of unobserved heterogeneity by comparing estimation from the mixed logit model, mixed logit model with heterogeneity in means, and mixed logit models with heterogeneity in means and variance (13). These studies identify the importance of taking heterogeneity effects into consideration when analyzing EVs' crash data. ...
... For PHEVs, an increase in occupant number is related to a decrease in the likelihood of both severe and light injuries. For HEVs and ICEVs, an increase in occupant number is associated with an increase in the likelihood of both severe and light injuries, which is consistent with the research result of Seraneeprakarn et al. (13). ...
With an increasing market penetration of electric vehicles (EVs) in the traffic mix, it become necessary to examine crashes involving EVs. In addition, there is a need to identify differences compared with traditional internal combustion engine vehicles (ICEVs), as EVs are heavier and have different performance characteristics than ICEVs. To date, there is limited research comparing crash characteristics among EVs and ICEVs and further, differentiating among different types of EVs: battery electric vehicles (BEVs), plug-in hybrid electric vehicles (PHEVs), and hybrid electric vehicles (HEVs). To fill this research gap, this paper estimates crash injury frequency and crash severity outcomes through statistical regression analyses. The statistical models and hypothesis testing results suggest both similarities and differences in crash characteristics among BEVs, PHEVs, HEVs, and ICEVs. The similarity lies in human-related factors and traffic-related factors, and the differences come from four types of factors including vehicle, roadway, crash, and environment. The potential reasons (in terms of vehicles’ engine type, software, and hardware) that could contribute to the differences in crash characteristics among four types of vehicles are discussed. The findings of this paper can provide insights into devising safety regulations for EVs. For example, EVs equipped with advanced driving assistant technologies can help relieve crash injury counts. However, the high acceleration rate of electric motors could positively contribute to the crash severity, and the front of BEVs needs more protection since head-on crashes of BEVs cause more severe crashes.
... However, it does not account for the potential influence of mean and variance heterogeneity on the random parameter (RP) trend. Consequently, an advanced version of the stochastic parameter LM, which considers both mean and variance heterogeneity, has been proposed (Yu, Zheng, and Ma 2020b;Alnawamsi and Mannering 2023;Behnood and Mannering 2017;Wang et al. 2022b;2022c;2022d;Seraneeprakarn et al. 2017). Yu, Zheng, and Ma (2020b) utilized this enhanced RPLM to analyze four years of rear-end crash data in North Carolina, acknowledging the model's superior fit compared to the ordinary LM. propsed the RPLM to analyze differences in day and night bicyclist injury severities. ...
... where P ij is the probability of the severity i in the j th crash; f (b|w) is the probability density function of the random vector b; w is the parameter vector describing the probability density; I is the set of all injury levels (property damage, minor injuries, and severe injuries). According to Seraneeprakarn et al. (2017), when considering the heterogeneity of mean and variance in the model, b i can be expressed as: ...
High fatality rates in frequent rear-end crashes have underscored significant safety concerns in China. This study aims to explore the mechanisms and determinants of rear-end crashes, with a particular focus on the factors influencing crash severity during weekdays and weekends (W-W). Employing the Random Parameter Logit Model (RPLM) to account for variability in data, we analyzed W-W rear-end crashes on the Beijing-Shanghai Expressway in Jiangsu province from 2017 to 2019. These crashes were classified into three severity levels: no injury, minor injury, and severe injury. Our comprehensive analysis covered variables from temporal, roadway, vehicle, crash, and environmental categories, alongside calculating the marginal effects of each significant variable on crash severity. Findings reveal temporal instability over the three-year period and notable differences in W-W crash severity. Out of all variables, four displayed random parameter characteristics, indicating potential interactions that influence crash outcomes. Specifically, our results indicate that rear-end crashes involving three or more vehicles on bridges are more likely to result in casualties. Interchange segments typically saw no injuries in two-vehicle crashes. Speeding during winter or on sunny days significantly increases the risk of injuries and fatalities. Furthermore, rear-end crashes in interchange areas during winter are particularly prone to causing injuries. These findings offer guidance for the development of effective safety countermeasures targed at different pediods.
... Huang et al. pointed out that the statistical analysis results could support strong heterogeneity effects in 1 crash data (6). Based on this conclusion, Seraneeprakarn et al. further validated the influence of 2 unobserved heterogeneity by comparing estimation from the mixed logit model, mixed logit model with 3 heterogeneity in means, and mixed logit models with heterogeneity in means and variance (7). These 4 studies identify the importance of taking heterogeneity effects into consideration when analyzing EVs' 5 crash data. ...
Powered by electric engines, electric vehicles (EVs) exhibit unique dynamic characteristics that may lead to different crash characteristics and outcomes compared with traditional internal combustion engine vehicles (ICEVs). This might be particularly true for vulnerable road users (VRUs), such as pedestrians and cyclists. Motivated by these concerns, this paper delves into the comparative analysis of crashes involving EVs and VRUs, exploring how crash characteristics and injury severities differ from those involving VRUs and ICEVs. Employing statistical testing and binary probit regression analyses, this study analyzes crash data from Chicago spanning from 2015 to 2022. Spatial and temporal constraints were applied to filter ICEV crashes, ensuring similar environmental conditions and VRU exposure for the considered crashes. Innovatively, this study supplements traditional police crash reports with Google Street View (GSV) images and employs neural network models to uncover previously unreported environmental variables at crash scenes. The results reveal both similarities and disparities in the characteristics of crash involved VRUs between EVs and ICEVs. However, significant differences in factors, such as VRU type (pedestrians or cyclists), hit-and-run incidents, damage level, crash hour, crash weekday, weather conditions, and road surface conditions, along with the influence of season and road surface condition on injury severity, were observed between EVs and ICEVs. These distinctions may be attributed to driver demographics, vehicle design, and usage characteristics. These insights can guide the development of safety regulations for EVs and aid in devising specific safety measures and policies for VRUs, including pedestrians and cyclists.
This study combines econometric and geospatial methods to analyze the impact of cellular coverage on distracted driving incidents, crashes, and injuries in Oregon from 2017 to 2020, focusing on cell phone-related crashes. Despite reduced travel in 2020, these crashes remained high. Geospatial tools identified urban hotspots like Portland and Salem. We used the mixed logit model to evaluate factors like driver demographics, vehicle characteristics, and environmental conditions, shedding light on the economic aspects of injury severity. Results highlight the crucial role of seatbelt use in reducing injury severity. The study underlines the need for comprehensive strategies to combat distracted driving in Oregon for better road safety and to lower economic costs associated with such incidents.
This study comprehensively investigated the factors affecting crash injury severity associated with commercial motor vehicles (CMVs), that is, large trucks and buses, on Interstate-65 (I-65) in the state of Kentucky. Unconventionally explored microscopic real-time weather variables (i.e., air temperature, relative humidity, solar radiation, wind speed, and precipitation) were extracted from 80 mesonet stations in Kentucky and used in the analysis. Other variables explored included hourly traffic volume and speed and driver-, roadway-, vehicle-, and environment-related predictors. Recent crashes within a 5-year, 4-month period (January 1, 2016, through April 30, 2021) along I-65 were used. The association rules mining (ARM) technique was used to uncover associations between real-time weather and CMV-related severe injuries (KA). The ARM analysis showed that severe CMV-related crashes were associated with the co-occurrence of the following real-time weather conditions: solar radiation ≤ 5 W/m ² , relative humidity 65% to 90%, and air temperature ≤ 50°F. Furthermore, the correlated mixed logit with heterogeneity in means (CMXLHM) model was applied to identify significant factors affecting CMV-related crash severity while accounting for unobserved heterogeneity and correlations among the random parameters. The CMXLHM model results showed that solar radiation ≤ 5 W/m ² and air temperature ≤ 50°F increased the probability of severe CMV crash injury outcome by 46% and 26.89%, respectively. Furthermore, middle-age drivers (31 to 59 years old), speeding, distracted driving, and weekend-related crashes were associated with increased CMV crash injury outcomes. Safety recommendations are proposed to enhance CMV safety. One example is feeding specific real-time weather states to dynamic message signs for safety alerts.
Pedestrian mobility has become an increasingly important concern in transportation system analysis because of its positive impacts on the environment and healthy lifestyles. However, pedestrian safety in a vehicle-dominated transportation network remains a concern and potential barrier to pedestrian mobility, with pedestrian intersection safety being of particular concern. In addition, it is important to understand how pedestrian safety has been affected by the COVID-19 pandemic, perhaps permanently shifting pedestrian injury risks. This research seeks to provide insight into how pedestrian injury risks at intersections have changed as a result of the pandemic by estimating a series pedestrian injury severity models. To do so, unconstrained and partially constrained random parameters multinomial logit models with heterogeneity in the means of random parameters were estimated. Using Florida data, two one-year periods (one year before and one year after the COVID-19 pandemic) were defined based on vehicle miles traveled. The sample includes 3,780 single pedestrian-single vehicle crashes (2,348 from daytime and 1,432 from nighttime). A broad range of variables was considered to assess how the parameters may have shifted between the before and after periods. A series of likelihood ratio tests were conducted to examine the stability of model parameter estimates across the predefined time periods as well as to determine the differences between the daytime and nighttime crash injury severity outcomes. The results show that the nighttime crashes experienced more temporal shifts compared to daytime crashes. The findings also showed that both pedestrian and driver behavior played key temporally-shifting roles before, during and after the COVID-19 pandemic period. Finally, the out-of-sample simulations suggest that pedestrian injuries have become more severe after the pandemic.
Ensuring the safety of students is a multifaceted task, extending beyond the mere design of vehicles. It involves analyzing crash data related to school buses, which plays a critical role in preventing injuries and fatalities of children both in and around these vehicles. The National Highway Traffic Safety Administration ¹ gathers comprehensive data on a wide range of school bus-related crashes, from minor incidents to serious and fatal crashes. The school bus crash data not only educates parents and caregivers but also supports various prevention campaigns. The objective of this study is to investigate the different risk factors that influence the severity of injuries in school bus crashes, considering variations in time and location. To achieve this, a Random thresholds random parameters ordered probit (RPOP) model is used, which is particularly effective in identifying and accounting for unobserved variables that might affect the outcomes. Moreover, the model results suggest that springtime and highway locations significantly impact crash outcomes, considering crash details, vehicle information, and driver-passenger characteristics. These findings enhance our comprehension of the diverse factors that contribute to school bus crashes. This improved understanding aids in the development of effective strategies to reduce the severity of injuries resulting from such crashes.
Road users (drivers, passengers, pedestrians, and Animals) are exposed to hazardous events during their commute. With 23 % of global fatalities among pedestrians, their safety continues to be a principal interest for policymakers worldwide. Owing to limited budgets available, there is a growing emphasis on data-driven stochastic models to decide on policies. However, statistical models have limitations due to crash data having redundant features, inherent heterogeneity, and unobserved characteristics. The random parameter model framework addresses the unobserved heterogeneity, but redundant features and inherent heterogeneity among the data's characteristics still compute the biased estimates. This is further complicated if the data has spatiotemporal attributes. To address this, we developed two visual hazardous (VH) models: (i) addresses the unobserved heterogeneity in the data, and (ii) addresses the dimensionality, inherent heterogeneity among the characteristics and unobserved heterogeneity in the collected data after spatiotemporal pattern identification. The feature selection model reduces the dimensionality, whereas latent class clustering classifies the data into maximum heterogeneity between classes. This integration reduces bias in the estimates. As a use-case, pedestrian crosswalk crashes for a decade (2009–2018) in the Indian state of Tamil Nadu extracted from the Road Accident Database Management System (RADMS) was used to understand model performance. This data comprises the crash location, road, vehicle, driver, pedestrian, and environment details. Results show that visual hazardous model 2 allows for generating crash scenarios with five homogeneous sub-classes and the magnitude with marginal effects of contributing factors impacting it. For example, pedestrians during their crosswalks are likely to sustain 82% more chance of fatal/grievous injuries on expressways (posted speed limit: 100 km per hour) in annual hazardous zone locations. Working pedestrian age group (25–64 years), an older pedestrian (>64 years), the pedestrian position on a pedestrian crossing and not in the centre of the road, pedestrian action: walking along the edge of the road, multiple lanes, two lanes, paved shoulder, straight and flat road, motorcycle, bus, truck, medium-duty vehicle, illegal driver (<=17 years), going ahead/ overtaking, high speed, expressways, and rural region were statistically significant (positively) contributing to the fatal/grievous injury pedestrian crashes during their crosswalk. This technique serves as a structure for engineers, researchers, and policymakers to formulate effective countermeasures that enhance road safety.
Contemporarily, hybrid vehicles have been increasingly accepted by consumers because of environmental, economic and nonrenewable (e.g. fossil fuels) resource concerns. In addition to the differences in fuel source, hybrid vehicles are different from the fuel-engine vehicles in terms of ride characteristics, ride noise, weight, etc. These differences can be sources of potential shifts in severity distributions involving hybrid vehicle crashes. However, how the attributes of a hybrid vehicle affect our inferences on crash severity propensities is a largely unaddressed issue in the literature. This paper attempts to contribute some insight into the impact of hybrid vehicle attributes on crash severity propensities by using a hierarchical mixed logit model to predict the crash severity in crashes involving hybrid vehicles. The hierarchical model allows for the identification of factors that can influence of the mean of the random parameters in the mixed logit. Using statewide data from reported crashes in Washington State involving hybrid vehicles, this paper develops the model for the severity of crashes involving hybrid vehicles, by considering factors such as roadway conditions, environment factors, occupant attributes and vehicle characteristics. The hierarchical influences consist of hybrid vehicle attributes, due to the fact that at least one of the vehicles involved in the crash dataset is a hybrid. The research results shows that the hierarchical mixed logit is a plausible approach for gaining insight into the particular impact of hybrid vehicle attributes on crash severity parameters.
This paper uses driving simulation data and surveys conducted in 2014 and 2015 in Buffalo, NY, to study the factors that affect perceived (self-reported, based on surveys) and observed (as measured, based on driving simulation experiments) aggressive driving behavior. Perceived and observed aggressive driving behavior are likely to share unobserved characteristics. To simultaneously account for this cross-equation error correlation, and for unobserved heterogeneity and panel data effects, a grouped random parameters bivariate probit model is estimated. The results control and account for a number of socio-demographic, driving experience and exposure, and behavioral and other characteristics. The findings reveal that different variables play in how aggressive driving behavior is perceived and observed, and the results imply that some drivers may perceive their driving behavior as non-aggressive when it is aggressive (or the opposite). The grouped random parameters bivariate probit model results are compared to their univariate probit, full information maximum likelihood bivariate probit, bivariate probit model with random effects, and random parameters bivariate probit model counterparts, and the results reveal the statistical superiority of the former, in terms of explanatory power, model fit, and forecasting accuracy.
This paper seeks to investigate the effects of passengers on driver-injury severities. Using single-vehicle crashes, a random parameters logit model with heterogeneity in parameter means is estimated to explore the differences in driver-injury severities in three distinct subgroups; vehicles with one occupant (driver-only), vehicles with two occupants (driver and a passenger), and vehicles with three occupants (driver and two passengers). In addition to considering age, gender and the presence of the passenger(s), a wide range of variables that potentially affect driver-injury severity were considered, including weather conditions, roadway characteristics, vehicle characteristics and driver attributes. Estimation results show that both the age and the gender of the passenger(s) significantly affected driver-injury severities. The findings of this research point toward the need to further study the potentially complex interactions between drivers and passengers.
The current study developed a simultaneous model of injury severity outcomes of all occupants in multi-vehicle crashes including all the drivers and the passengers of all vehicles involved in a crash. Specifically, a Modified Rank Ordered Logit (MROL) methodology that can predict the relative order of occupant injury severity as well as the actual injury severity was developed. The final model captures the effects of several key occupant, vehicle, and accident level variables on four possible levels of injury severity. The results indicate the presence of accident-specific unobserved factors that influence the severity outcomes of all people involved in the crash as well as unobserved heterogeneity in the effect of key covariates including occupant’s gender and speed limit. The performance of the MROL model was compared with the traditional mixed multinomial logit (MMNL) model that is the most commonly used model for injury severity analysis. Overall, the results demonstrate superior predictive ability of the MROL model in comparison to the MMNL model. The traditional MMNL model performed satisfactory in terms of replicating the simple shares of different injury severity levels across all occupants. However, the performance of the MMNL model dropped significantly when the observed and predicted shares were compared for combinations of injury severity levels among crashes involving multiple occupants. Lastly, elasticity effects were computed to demonstrate considerably different policy implications of the MROL and MMNL models.
Several recent transportation safety studies have indicated the importance of accounting for correlated
outcomes, for example, among different crash types, including differing injury-severity levels. In this
paper, we discuss inference for such data by introducing a flexible Bayesian multivariate model. In particular,
we use a Dirichlet process mixture to keep the dependence structure unconstrained, relaxing the
usual homogeneity assumptions. The resulting model collapses into a latent class multivariate model that
is in the form of a flexible mixture of multivariate normal densities for which the number of mixtures
(latent components) not only can be large but also can be inferred from the data as part of the analysis.
Therefore, besides accounting for correlation among crash types through a heterogeneous correlation
structure, the proposed model helps address unobserved heterogeneity through its latent class component.
To our knowledge, this is the first study to propose and apply such a model in the transportation
literature. We use the model to investigate the effects of various factors such as built environment characteristics
on pedestrian and cyclist injury counts at signalized intersections in Montreal, modeling both
outcomes simultaneously. We demonstrate that the homogeneity assumption of the standard multivariate
model does not hold for the dataset used in this study. Consequently, we show how such a spurious
assumption affects predictive performance of the model and the interpretation of the variables based on
marginal effects. Our flexible model better captures the underlying complex structure of the correlated
data, resulting in a more accurate model that contributes to a better understanding of safety correlates
of non-motorist road users. This in turn helps decision-makers in selecting more appropriate countermeasures
targeting vulnerable road users, promoting the mobility and safety of active modes of
transportation.
The cornerstone of transportation infrastructure asset management is managing the physical infrastructure, with pavement preservation being one of the most critical and costly assets. Preserving pavements in an appropriate manner extends their service life, and most importantly improves motorists’ safety and satisfaction while saving public tax dollars. To that end, this paper presents a methodology to estimate pavement performance thresholds that are cost-effective and safe for users. Using data from Indiana, the relationships of the three criteria, i.e., safety (accident rates), normalized treatment cost and pavement service life, with the pavement performance (roughness, rutting, overall rating, and surface deflection), road geometry, traffic characteristics and climate - are investigated and estimated. These relationships are utilized in a multi-objective optimization and goal-programming scheme to identify performance threshold values that trigger preservation treatments. These analytically determined threshold values are found to be comparable to historical thresholds and thresholds derived from experts’ and users’ opinions.
The link between risk-taking behavior in various domains of life has long been an area of debate among economists and psychologists. Using an extensive data set from Denmark, this study provides an empirical investigation of the link between risky driving and risk taking in other aspects of life, including risk-taking behavior in financial and labor-market decisions. Specifically, we establish significant positive correlations between individuals’ risk-taking behavior in car driving and their risk-taking behavior in financial and labor-market decisions. However, we find that the strength of these correlations vary significantly between genders, and across risk decisions. These correlations and their differences across genders get stronger when we construct more ‘‘homogeneous’’ groups by restricting our sample to those individuals with at least some stock-market participation. Overall, the empirical results in this study suggest that risk-taking behavior in various aspects of life can be associated, and our results corroborate previous evidence on the link between individuals’ risk preferences across various aspects of life. This implies that individuals’ driving behavior, which is commonly unobservable, can be more fully understood using observable labor market and financial decisions of individuals.
This study explores the differences in pedestrian injury severity in three distinct economic time periods from the recent global recession (the Great Recession): pre-recession, recession, and post-recession. Using data from pedestrian crashes in Chicago, Illinois over an eight-year period, separate time-period models of pedestrian-injury severities (with possible outcomes of severe injury, moderate injury, and minor injury) were estimated using latent-class logit and mixed logit models. Likelihood ratio tests were conducted to examine the overall stability of model estimates across time periods and marginal effects of each explanatory variable were also considered to investigate the temporal stability of the effect of individual parameter estimates on pedestrian injury-severity probabilities. A wide range of variables potentially affecting injury severities was considered including time, location, and severity of crashes, as well as data on roadway and environmental conditions, pedestrian characteristics, and crash characteristics. Our findings show significant temporal instability, which likely results from a combination of the economic recession and the long-term evolution of the influence of factors that affect pedestrian-injury severity. Understanding and explicitly modeling the evolution of driver and pedestrian behavior is a promising direction for future research, but this would unfortunately require far more extensive data than is currently available in traditional safety databases.
This paper uses data collected over a five-year period between 2005 and 2009 in Indiana to estimate random parameters multivariate tobit and zero-inflated count data models of accident injury-severity rates and frequencies, respectively. The proposed modeling approach accounts for unobserved factors that may vary systematically across segments with and without observed or reported accident injury-severities, thus addressing unobserved, zero-accident state and non-zero-accident state heterogeneity. Moreover, the multivariate setting allows accounting for contemporaneous cross-equation error correlation for modeling accident injury-severity rates and frequencies as systems of seemingly unrelated equations. The tobit and zero-inflated count data modeling approaches address the excessive amount of zeros inherent in the two sets of dependent variables (accident injury-severity rates and frequencies, respectively), which are – in nature – continuous and discrete count data, respectively, that are left-censored with a clustering at zero. The random parameters multivariate tobit and zero-inflated count data models are counter-imposed with their equivalent fixed parameters and lower order models, and the results illustrate the statistical superiority of the presented models. Finally, the relative benefits of random parameters modeling are explored by demonstrating the forecasting accuracy of the random parameters multivariate models with the software-generated mean s of the random parameters, and with the observation-specific s of the random parameters.