Identifying crash-prone locations with quantile regression

Application of Quantile Mixed Model for modeling Traffic Barrier Crash Cost

Run-off the road crashes account for a significant proportion of severe injuries to vehicle occupants. Traffic barriers have been installed with an objective to keep vehicles on the roadway, and prevent them from hitting natural obstacles like trees or boulders. However, still injuries and fatalities of barrier crashes account for high proportion of fatalities on roadway. Due to challenging geometrics characteristics of Wyoming's roadway, a high mileage of barriers has been installed in the state. The high mileages of barriers result in a high number of barrier crashes in terms of crash frequency and severity due to high exposure. Previous studies mainly focused on crash frequency or individual crash severity. However, it has been recognized the importance of accounting for both aspects of crash severity, and crash frequency. So, in this study, crashes are aggregated across different barriers, and those crashes were converted into costs by considering the impacts of both crash severity and frequency. However, one of the main challenges of this type of dataset is highly skewness of crash data due to its sparseness nature. An improper use of model distribution of crash cost would result in biased estimations of the covariates, and erroneous results. Thus, in order to address this issue, a semi-parametric method of quantile regression technique was implemented to account for the skewness of the response by relaxing model distribution parameters. Also, to account for the heterogeneity in the dataset due to barriers' types, a random intercept model accounting for the structure of the data was implemented. In addition, interaction terms between significant predictors were considered. Understanding what factors with which magnitude contribute to the barrier crash costs is crucial for the future barriers' optimization process. Thus, contributory factors to barriers crash cost with high, medium, and low values, corresponding to 95^th, 70^th, and 60^th percentiles were considered, and a comparison was made across these models. It was found, for instance, that although factors such as rollover, driving under the influence, and presence of heavy truck all have contributory impacts on the cost of crashes, their impacts are greater on higher quantiles, or higher barriers' costs. These models were compared from various perspectives such as intra class correlation (ICC), and standard error of coefficients. This study highlights the changes in coefficient estimates while modeling crash costs.

Safety sensitivity to roadway characteristics: A comparison across highway classes

This paper examined the accident risk factors associated with highway traffic and roadway design, for each of three highway classes in the United States using a bivariate modeling framework involving two levels of accident severity. With regard to the highest class (Interstates), the results suggest that, compared to no-casualty accidents, casualty accidents are more sensitive to traffic volume and average vertical grade, but less sensitive to the inside shoulder width and the median width. For US Roads, it was determined that, compared to no-casualty accidents, casualty accidents are more sensitive to traffic volume, outside shoulder width, pavement condition, and median width but less sensitive to the average vertical grade. For the relatively lowest-class roads (State Roads), it was determined that, compared to no-casualty accidents, casualty accidents are more sensitive to the traffic volume, lane width, outside shoulder width, and pavement condition. Compared to the relatively lower-class highways, accidents at higher-class highways are more sensitive to: changes in traffic volume, average vertical grade, median width, inside shoulder width, and the pavement condition (no-casualty accidents only); but less sensitive to changes in lane width, pavement condition (casualty accidents only), and the outside shoulder width. This variation in sensitivity across the different road classes could be attributed to the differences in road geometry standards across the road classes, as the results seem to support the hypothesis that these standards strongly influence accident occurrence. It is hoped that the developed bivariate negative binomial models can help highway engineers to evaluate their current design standards and policy, and to assess the safety consequences of changes in these standards in each road class.

Identifying the safety factors over traffic signs in state roads using a panel quantile regression approach

OBJECTIVE

This study intended to investigate the interactions between accident rate and traffic signs on state roads located in Croatia and accommodate the heterogeneity attributed to unobserved factors. Data from 130 state roads between 2012 and 2016 were collected from Traffic Accident Database System maintained by the Republic of Croatia's Ministry of the Interior.

METHODS

To address the heterogeneity, a panel quantile regression model was proposed, in which a quantile regression model offers a more complete view and a highly comprehensive analysis of the relationship between accident rate and traffic signs, and the panel data model accommodates the heterogeneity attributed to unobserved factors.

RESULTS

Results revealed that (1) low visibility of material damage (MD) and death or injury (DI) increased the accident rate; (2) the number of mandatory signs and the number of warning signs were more likely to reduce the accident rate; (3) the average speed limit and the number of invalid traffic signs per kilometer exhibited a high accident rate.

CONCLUSIONS

To our knowledge, this study is the first attempt to analyze the interactions between accident consequences and traffic signs by employing a panel quantile regression model; by including visibility, the present study demonstrates that low visibility causes a relatively higher risk of MD and DI. It is noteworthy that average speed limit positively corresponds with accident rate; the number of mandatory signs and the number of warning signs are more likely to reduce the accident rate; and the number of invalid traffic signs per kilometer is significant for the accident rate; thus, regular maintenance should be performed for a safer roadway environment.

Applying quantile regression for modeling equivalent property damage only crashes to identify accident blackspots

Hot spot identification (HSID) aims to identify potential sites-roadway segments, intersections, crosswalks, interchanges, ramps, etc.-with disproportionately high crash risk relative to similar sites. An inefficient HSID methodology might result in either identifying a safe site as high risk (false positive) or a high risk site as safe (false negative), and consequently lead to the misuse the available public funds, to poor investment decisions, and to inefficient risk management practice. Current HSID methods suffer from issues like underreporting of minor injury and property damage only (PDO) crashes, challenges of accounting for crash severity into the methodology, and selection of a proper safety performance function to model crash data that is often heavily skewed by a preponderance of zeros. Addressing these challenges, this paper proposes a combination of a PDO equivalency calculation and quantile regression technique to identify hot spots in a transportation network. In particular, issues related to underreporting and crash severity are tackled by incorporating equivalent PDO crashes, whilst the concerns related to the non-count nature of equivalent PDO crashes and the skewness of crash data are addressed by the non-parametric quantile regression technique. The proposed method identifies covariate effects on various quantiles of a population, rather than the population mean like most methods in practice, which more closely corresponds with how black spots are identified in practice. The proposed methodology is illustrated using rural road segment data from Korea and compared against the traditional EB method with negative binomial regression. Application of a quantile regression model on equivalent PDO crashes enables identification of a set of high-risk sites that reflect the true safety costs to the society, simultaneously reduces the influence of under-reported PDO and minor injury crashes, and overcomes the limitation of traditional NB model in dealing with preponderance of zeros problem or right skewed dataset.

Analysis of U.S. freight-train derailment severity using zero-truncated negative binomial regression and quantile regression

Derailments are the most common type of freight-train accidents in the United States. Derailments cause damage to infrastructure and rolling stock, disrupt services, and may cause casualties and harm the environment. Accordingly, derailment analysis and prevention has long been a high priority in the rail industry and government. Despite the low probability of a train derailment, the potential for severe consequences justify the need to better understand the factors influencing train derailment severity. In this paper, a zero-truncated negative binomial (ZTNB) regression model is developed to estimate the conditional mean of train derailment severity. Recognizing that the mean is not the only statistic describing data distribution, a quantile regression (QR) model is also developed to estimate derailment severity at different quantiles. The two regression models together provide a better understanding of train derailment severity distribution. Results of this work can be used to estimate train derailment severity under various operational conditions and by different accident causes. This research is intended to provide insights regarding development of cost-efficient train safety policies.

Modeling vehicle operating speed on urban roads in Montreal: a panel mixed ordered probit fractional split model

Vehicle operating speed measured on roadways is a critical component for a host of analysis in the transportation field including transportation safety, traffic flow modeling, roadway geometric design, vehicle emissions modeling, and road user route decisions. The current research effort contributes to the literature on examining vehicle speed on urban roads methodologically and substantively. In terms of methodology, we formulate a new econometric model framework for examining speed profiles. The proposed model is an ordered response formulation of a fractional split model. The ordered nature of the speed variable allows us to propose an ordered variant of the fractional split model in the literature. The proposed formulation allows us to model the proportion of vehicles traveling in each speed interval for the entire segment of roadway. We extend the model to allow the influence of exogenous variables to vary across the population. Further, we develop a panel mixed version of the fractional split model to account for the influence of site-specific unobserved effects. The paper contributes substantively by estimating the proposed model using a unique dataset from Montreal consisting of weekly speed data (collected in hourly intervals) for about 50 local roads and 70 arterial roads. We estimate separate models for local roads and arterial roads. The model estimation exercise considers a whole host of variables including geometric design attributes, roadway attributes, traffic characteristics and environmental factors. The model results highlight the role of various street characteristics including number of lanes, presence of parking, presence of sidewalks, vertical grade, and bicycle route on vehicle speed proportions. The results also highlight the presence of site-specific unobserved effects influencing the speed distribution. The parameters from the modeling exercise are validated using a hold-out sample not considered for model estimation. The results indicate that the proposed panel mixed ordered probit fractional split model offers promise for modeling such proportional ordinal variables.

Understanding socioeconomic and racial differences in adult lung function

RATIONALE

The contribution of socioeconomic factors to racial differences in the distribution of lung function is not well understood.

OBJECTIVES

We investigated the contribution of socioeconomic factors to racial differences in FEV₁ using statistical tools that allow for examination across the population distribution of FEV₁.

METHODS

We compared FEV₁ for white and African-American participants (aged 20-80 yr) in NHANES III with greater than or equal to two acceptable maneuvers to a restricted sample following the routine exclusion criteria used to derive population reference equations. Ordinary least squares and quantile regression analyses using spirometric, anthropometric, and socioeconomic data (high school completion) were performed separately by sex for both data sets.

MEASUREMENTS AND MAIN RESULTS

In the entire sample with acceptable spirometry (n ¼ 9,658), high school completion was associated with a mean 69.13-ml increase in FEV₁ for males (P , 0.05) and a mean 50.75-ml increase in FEV₁ for females (P , 0.01). In quantile regression analysis, we observed a significant racial difference in the association of high school completion with FEV₁ among both sexes that varied across the distribution; college completion was associated with an additional increase in FEV₁ for white males (70.36-250.76 ml) and white females (57.87-317.77 ml). Routine exclusion criteria differentially excluded individuals by age, race, and education. In the restricted sample (n ¼ 2,638), the association with high school completion was not significant.

CONCLUSIONS

High school completion is associated with racially patterned improvements in the FEV₁ of adults in the general population. The application of routine exclusion criteria leads to underestimates of the role of high school completion on FEV₁.