Venkataraman N. Shankar

Effect of roadway geometrics and environmental factors on rural freeway accident frequencies.

This paper explores the frequency of occurrence of highway accidents on the basis of a multivariate analysis of roadway geometrics (e.g. horizontal and vertical alignments), weather, and other seasonal effects. Based on accident data collected in the field, a negative binomial model of overall accident frequencies is estimated along with models of the frequency of specific accident types. Interactions between weather and geometric variables are proposed as part of the model specifications. The results of the analysis uncover important determinants of accident frequency. By studying the relationship between weather and geometric elements, this paper offers insight into potential measures to counter the adverse effects of weather on highway sections with challenging geometrics.

Highway accident severities and the mixed logit model: An exploratory empirical analysis

Many transportation agencies use accident frequencies, and statistical models of accidents frequencies, as a basis for prioritizing highway safety improvements. However, the use of accident severities in safety programming has been often been limited to the locational assessment of accident fatalities, with little or no emphasis being placed on the full severity distribution of accidents (property damage only, possible injury, injury)-which is needed to fully assess the benefits of competing safety-improvement projects. In this paper we demonstrate a modeling approach that can be used to better understand the injury-severity distributions of accidents on highway segments, and the effect that traffic, highway and weather characteristics have on these distributions. The approach we use allows for the possibility that estimated model parameters can vary randomly across roadway segments to account for unobserved effects potentially relating to roadway characteristics, environmental factors, and driver behavior. Using highway-injury data from Washington State, a mixed (random parameters) logit model is estimated. Estimation findings indicate that volume-related variables such as average daily traffic per lane, average daily truck traffic, truck percentage, interchanges per mile and weather effects such as snowfall are best modeled as random-parameters-while roadway characteristics such as the number of horizontal curves, number of grade breaks per mile and pavement friction are best modeled as fixed parameters. Our results show that the mixed logit model has considerable promise as a methodological tool in highway safety programming.

Evaluating Median Crossover Likelihoods with Clustered Accident Counts: An Empirical Inquiry Using the Random Effects Negative Binomial Model

Insights into plausible methodological frameworks specifically with respect to two key issues—(1) mathematical formulation of the underlying process affecting median crossover accidents and (2) the factors affecting median crossover frequencies in Washington State—are provided in this study. Random effects negative binomial (RENB) and the cross-sectional negative binomial (NB) models are examined. The specification comparisons indicate benefits from using the RENB model only when spatial and temporal effects are totally unobserved. When spatial and temporal effects are explicitly included, the NB model is statistically adequate, while the RENB model appears to lose its distributional advantage. Such findings might be artifacts of the median crossover accident dataset used in this study. While the NB model appears to be the superior model in the present case of median crossover accidents, the marginally inferior performance of the RENB model warrants further examination through application to regular accident types in light of its flexibility to incorporate temporal and cross-sectional variations simultaneously in panel counts. From a predictive standpoint, RENB offers advantages in terms of model transferability and updating.

Age and pedestrian injury severity in motor-vehicle crashes : A heteroskedastic logit analysis

This research explores the injury severity of pedestrians in motor-vehicle crashes. It is hypothesized that the variance of unobserved pedestrian characteristics increases with age. In response, a heteroskedastic generalized extreme value model is used. The analysis links explanatory factors with four injury outcomes: fatal, incapacitating, non-incapacitating, and possible or no injury. Police-reported crash data between 1997 and 2000 from North Carolina, USA, are used. The results show that pedestrian age induces heteroskedasticity which affects the probability of fatal injury. The effect grows more pronounced with increasing age past 65. The heteroskedastic model provides a better fit than the multinomial logit model. Notable factors increasing the probability of fatal pedestrian injury: increasing pedestrian age, male driver, intoxicated driver (2.7 times greater probability of fatality), traffic sign, commercial area, darkness with or without streetlights (2-4 times greater probability of fatality), sport-utility vehicle, truck, freeway, two-way divided roadway, speeding-involved, off roadway, motorist turning or backing, both driver and pedestrian at fault, and pedestrian only at fault. Conversely, the probability of a fatal injury decreased: with increasing driver age, during the PM traffic peak, with traffic signal control, in inclement weather, on a curved roadway, at a crosswalk, and when walking along roadway.

Underreporting in traffic accident data, bias in parameters and the structure of injury severity models

Injury severities in traffic accidents are usually recorded on ordinal scales, and statistical models have been applied to investigate the effects of driver factors, vehicle characteristics, road geometrics and environmental conditions on injury severity. The unknown parameters in the models are in general estimated assuming random sampling from the population. Traffic accident data however suffer from underreporting effects, especially for lower injury severities. As a result, traffic accident data can be regarded as outcome-based samples with unknown population shares of the injury severities. An outcome-based sample is overrepresented by accidents of higher severities. As a result, outcome-based samples result in biased parameters which skew our inferences on the effect of key safety variables such as safety belt usage. The pseudo-likelihood function for the case with unknown population shares, which is the same as the conditional maximum likelihood for the case with known population shares, is applied in this study to examine the effects of severity underreporting on the parameter estimates. Sequential binary probit models and ordered-response probit models of injury severity are developed and compared in this study. Sequential binary probit models assume that the factors determining the severity change according to the level of the severity itself, while ordered-response probit models assume that the same factors correlate across all levels of severity. Estimation results suggest that the sequential binary probit models outperform the ordered-response probit models, and that the coefficient estimates for lap and shoulder belt use are biased if underreporting is not considered. Mean parameter bias due to underreporting can be significant. The findings show that underreporting on the outcome dimension may induce bias in inferences on a variety of factors. In particular, if underreporting is not accounted for, the marginal impacts of a variety of factors appear to be overestimated. Fixed objects and environmental conditions are overestimated in their impact on injury severity, as is the effect of separate lap and shoulder belt use. Combined lap and shoulder belt usage appears to be unaffected. The parameter bias is most pronounced when underreporting of possible injury accidents in addition to property damage only accidents is taken into account.

Modeling crashes involving pedestrians and motorized traffic

This paper presents an empirical inquiry into the predictive modeling of crashes involving pedestrians and motorized traffic on roadways. Empirical models based on the negative binomial distribution and mixing distributions, such as the zero-inflated Poisson distribution, are presented and discussed in terms of their applicability to pedestrian crash phenomena. Key modeling issues relating to the presence of excess zeros as well as unobserved heterogeneity in pedestrian crash distributions are addressed. The empirical results show that zero-inflated count distributions, such as the zero-inflated Poisson, are promising methodologies for providing explanatory insights into the causality behind pedestrian-traffic crashes.

Driver-injury severity in single-vehicle crashes in California: A mixed logit analysis of heterogeneity due to age and gender

This research develops a mixed logit model of driver-injury severity in single-vehicle crashes in California. The research especially considers the heterogeneous effects of age and gender. Older drivers (65+ years old) were found to have a random parameter with about half the population having a higher probability of a fatal injury given a crash than the comparison group of 25-64 year olds with all other factors than age kept constant. The other half of the 65+ population had a lower probability of fatal injury. Heterogeneity was also noted in vehicle age, but related to the gender of the driver, with males linked to, on average, a higher probability of fatal injury in a newer vehicle compared with females, all other factors kept constant. These effects lend support to the use of mixed logit models in injury severity research and show age and gender based population heterogeneity. Several other factors were found to significantly increase the probability of fatal injury for drivers in single-vehicle crashes, most notably: male driver, drunk driving, unsafe speed, older driver (65+) driving an older vehicle, and darkness without streetlights.

Accident count model based on multiyear cross-sectional roadway data with serial correlation

The use of the negative multinomial model to form a predictive model of median crossover accident frequencies with a multiyear panel of crosssectional roadway data with a roadway section-specific serial correlation across time was explored. The negative multinomial model specification is compared with previous research, which used the same database but which also used negative binomial and random-effects negative binomial count models. If there is no section-specific correlation in the panel, the negative multinomial model becomes equivalent to the negative binomial. The differences in the estimation results between those models show that such a correlation exists in the data. The results show that the negative multinomial significantly outperforms the negative binomial and the random-effects negative binomial in terms of fit, with a statistically significantly higher likelihood at convergence. The signs of the coefficients were similar in all models; when the signs differed, the negative multinomial model results were more intuitive. Overall, the analysis supports the use of the negative multinomial count model to estimate median crossover accident frequency models that are based on panel data.

Marginal Impacts of Design, Traffic, Weather, and Related Interactions on Roadside Crashes

A multivariate model that incorporates the effects of design, traffic, weather, and related interactions with design variables on reported roadside crashes is presented. By providing for a framework that accounts for all measurable effects, the model minimizes the impact of omitted variable effects. Furthermore, the presented framework accounts for partial observability effects that stem from fluctuations in environmental conditions as well as unobserved effects that contribute to heterogeneity in the traffic safety network. A sample of 318 sections 1 mi long was used for the study. These sections represent the state highway network in the state of Washington on the basis of environmental and road classification factors and therefore were used for the collection of detailed precipitation, snowfall, and temperature data in addition to roadway and roadside design and traffic parameters. The resulting model suggests that the marginal impact of weather is both in main effects and interactive form, and that even after controlling for unobserved heterogeneity and partial observability, weather effects play a statistically significant role in roadside crash occurrence. In particular, it was found that in addition to precipitation, average monthly snowfall exceeding 4 in. and interactions between snow depths and horizontal curves were found to have a statistically significant effect on roadside crash frequency probabilities. The marginal effects of these variables were also statistically significant; furthermore, the contribution of weather and related interactions to the likelihood of roadside crash frequencies was approximately 19%, design main effects contributed to 33%, and traffic and design interactions contributed to 6%. Weather interactions with design contributed to approximately 6% of the overall likelihood. Traffic as a main effect contributed 36% to the overall roadside crash likelihood.

Children's Travel to School: Discrete Choice Modeling of Correlated Motorized and Nonmotorized Transportation Modes Using Covariance Heterogeneity

Childrens school travel mode is changing, especially away from walking and bicycling and towards private automobiles. Simultaneously we see warning signs from a public health standpoint as children are becoming less active. It has been suggested that walking and bicycling to or from school could help shift this trend, moving it towards greater activity, and researchers are therefore exploring choices of school-trip mode in relation to the pedestrian friendliness of the built environment. Mode-choice models are generally framed as multinomial logit (MNL) models. However, the limitations of MNL models can cause unrealistic effects when walking and bicycling are included with motorized modes. In this paper the focus is on accounting for individual-specific heterogeneity, since different children or families may have very different tastes or tolerances, such as travel time, when it comes to choosing between driving a private automobile, taking the school bus, bicycling, or walking to or from school. The results show that such heterogeneity exists, and that it is more important for nonmotorized modes than for the motorized modes. The results show that accounting for correlation across modes leads to more realistic marginal rates of substitution (cross-elasticities) across modes—in particular, an increase in the walking distance negatively affects the probability both of walking and of bicycling.