Pengpeng Xu

Modeling crash spatial heterogeneity: Random parameter versus geographically weighting

The widely adopted techniques for regional crash modeling include the negative binomial model (NB) and Bayesian negative binomial model with conditional autoregressive prior (CAR). The outputs from both models consist of a set of fixed global parameter estimates. However, the impacts of predicting variables on crash counts might not be stationary over space. This study intended to quantitatively investigate this spatial heterogeneity in regional safety modeling using two advanced approaches, i.e., random parameter negative binomial model (RPNB) and semi-parametric geographically weighted Poisson regression model (S-GWPR). Based on a 3-year data set from the county of Hillsborough, Florida, results revealed that (1) both RPNB and S-GWPR successfully capture the spatially varying relationship, but the two methods yield notably different sets of results; (2) the S-GWPR performs best with the highest value of Rd(2) as well as the lowest mean absolute deviance and Akaike information criterion measures. Whereas the RPNB is comparable to the CAR, in some cases, it provides less accurate predictions; (3) a moderately significant spatial correlation is found in the residuals of RPNB and NB, implying the inadequacy in accounting for the spatial correlation existed across adjacent zones. As crash data are typically collected with reference to location dimension, it is desirable to firstly make use of the geographical component to explore explicitly spatial aspects of the crash data (i.e., the spatial heterogeneity, or the spatially structured varying relationships), then is the unobserved heterogeneity by non-spatial or fuzzy techniques. The S-GWPR is proven to be more appropriate for regional crash modeling as the method outperforms the global models in capturing the spatial heterogeneity occurring in the relationship that is model, and compared with the non-spatial model, it is capable of accounting for the spatial correlation in crash data.

Sensitivity analysis in the context of regional safety modeling: Identifying and assessing the modifiable areal unit problem

A wide array of spatial units has been explored in current regional safety analysis. Since traffic crashes exhibit extreme spatiotemporal heterogeneity which has rarely been a consideration in partitioning these zoning systems, research based on these areal units may be subjected to the modifiable areal unit problem (MAUP). This study attempted to conduct a sensitivity analysis to quantitatively investigate the MAUP effect in the context of regional safety modeling. The emerging regionalization method-RECDAP (regionalization with dynamically constrained agglomerative clustering and partitioning) was employed to aggregate 738 traffic analysis zones in the county of Hillsborough to 14 zoning schemes at an incremental step-size of 50 zones based on spatial homogeneity of crash risk. At each level of aggregation, a Bayesian Poisson lognormal model and a Bayesian spatial model were calibrated to explain observed variations in total/severe crash counts given a number of zone-level factors. Results revealed that as the number of zones increases, the spatial autocorrelation of crash data increases. The Bayesian spatial model outperforms the Bayesian Poisson-lognormal model in accurately accounting for spatial autocorrelation effects, unbiased parameter estimates, and model performance, especially in cases with higher disaggregated levels. Zoning schemes with higher number of zones tend to have increasing number of significant variables, more stable coefficient estimation, smaller standard error, whereas worse model performance. The variables of population density and median household income show consistently significant effects on crash risk and are robust to variation in data aggregation. The MAUP effects may be significantly reduced if we just maintain at about 50% of the original number of zones (350 or larger). The present study highlights MAUP that is generally ignored by transportation safety analysts, and provides insights into the nature of parameter sensitivity to data aggregation in the context of regional safety modeling.

Revisiting crash spatial heterogeneity: A Bayesian spatially varying coefficients approach.

This study was performed to investigate the spatially varying relationships between crash frequency and related risk factors. A Bayesian spatially varying coefficients model was elaborately introduced as a methodological alternative to simultaneously account for the unstructured and spatially structured heterogeneity of the regression coefficients in predicting crash frequencies. The proposed method was appealing in that the parameters were modeled via a conditional autoregressive prior distribution, which involved a single set of random effects and a spatial correlation parameter with extreme values corresponding to pure unstructured or pure spatially correlated random effects. A case study using a three-year crash dataset from the Hillsborough County, Florida, was conducted to illustrate the proposed model. Empirical analysis confirmed the presence of both unstructured and spatially correlated variations in the effects of contributory factors on severe crash occurrences. The findings also suggested that ignoring spatially structured heterogeneity may result in biased parameter estimates and incorrect inferences, while assuming the regression coefficients to be spatially clustered only is probably subject to the issue of over-smoothness.

The effect of road network patterns on pedestrian safety: A zone-based Bayesian spatial modeling approach

Pedestrian safety is increasingly recognized as a major public health concern. Extensive safety studies have been conducted to examine the influence of multiple variables on the occurrence of pedestrian-vehicle crashes. However, the explicit relationship between pedestrian safety and road network characteristics remains unknown. This study particularly focused on the role of different road network patterns on the occurrence of crashes involving pedestrians. A global integration index via space syntax was introduced to quantify the topological structures of road networks. The Bayesian Poisson-lognormal (PLN) models with conditional autoregressive (CAR) prior were then developed via three different proximity structures: contiguity, geometry-centroid distance, and road network connectivity. The models were also compared with the PLN counterpart without spatial correlation effects. The analysis was based on a comprehensive crash dataset from 131 selected traffic analysis zones in Hong Kong. The results indicated that higher global integration was associated with more pedestrian-vehicle crashes; the irregular pattern network was proved to be safest in terms of pedestrian crash occurrences, whereas the grid pattern was the least safe; the CAR model with a neighborhood structure based on road network connectivity was found to outperform in model goodness-of-fit, implying the importance of accurately accounting for spatial correlation when modeling spatially aggregated crash data.

Injury Severity of Motorcycle Riders Involved in Traffic Crashes in Hunan, China: A Mixed Ordered Logit Approach

Issues related to motorcycle safety in China have not received enough research attention. As such, the causal relationship between injury outcomes of motorcycle crashes and potential risk factors remains unknown. This study intended to investigate the injury risk of motorcyclists involved in road traffic crashes in China. To account for the ordinal nature of response outcomes and unobserved heterogeneity, a mixed ordered logit model was employed. Given that the crash occurrence process is different between intersections and non-intersections, separate models were developed for these locations to independently estimate the impacts of various contributing factors on motorcycle riders’ injury severity. The analysis was based on the police-reported crash dataset obtained from the Traffic Administration Bureau of Hunan Provincial Public Security Ministry. Factors associated with a substantially higher probability of fatalities and severe injuries included motorcycle riders older than 60 years, the absence of helmets, motorcycle riders identified to be equal duty, and when a motorcycle collided with a heavy vehicle during the night time without lighting. Crashes occurred along county roads with curve and slope alignment or at regions with higher GDP were associated with an elevated risk of fatality of motorcycle riders, while unsignalized intersections were related to less severe injuries. Findings of this study are beneficial in forming several targeted countermeasures for motorcycle safety in China, including designing roads with appropriate road delineation and street lighting, strict enforcement for speeding and red light violations, promoting helmet usage, and improving the conspicuity of motorcyclists.

The influence of zonal configurations on macro-level crash modeling

ABSTRACT This study investigated the impacts of zonal configurations on macro-level traffic safety analysis for crashes of different severity levels. Bayesian multivariate Poisson-lognormal models with multivariate conditional auto-regressive priors were developed to account for the spatial autocorrelation between adjacent geographical units and correlations among crash types of four ordinal severity levels, i.e. fatality, severe injury, slight injury and no injury. For the purpose of evaluating the effects of zonal configurations on macro-level traffic safety analysis, the proposed model was calibrated using crash data of four types of geographical units, i.e. block group, traffic analysis zone, census tract and zip code tabulation area, in Hillsborough County of Florida. The study empirically revealed the extensive presence and the significance of MAUP in macro-level safety analysis based on the existing zonal configurations. It gave out a warning and encouraged more research efforts on rational application of macroscopic safety analysis with different zonal configurations.

Bayesian approach to model pedestrian crashes at signalized intersections with measurement errors in exposure

This study intended to identify the potential factors contributing to the occurrence of pedestrian crashes at signalized intersections in a densely populated city, based on a comprehensive dataset of 898 pedestrian crashes at 262 signalized intersections during 2010-2012 in Hong Kong. The detailed geometric design, traffic characteristics, signal control, built environment, along with the vehicle and pedestrian volumes were elaborately collected. A Bayesian measurement errors model was introduced as an alternative method to explicitly account for the uncertainties in volume data. To highlight the role played by exposure, models with and without pedestrian volume were estimated and compared. The results indicated that the omission of pedestrian volume in pedestrian crash frequency models would lead to reduced goodness-of-fit, biased parameter estimates, and incorrect inferences. Our empirical analysis demonstrated the existence of moderate uncertainties in pedestrian and vehicle volumes. Six variables were found to have a significant association with the number of pedestrian crashes at signalized intersections. The number of crossing pedestrians, the number of passing vehicles, the presence of curb parking, and the presence of ground-floor shops were positively related with pedestrian crash frequency, whereas the presence of playgrounds near intersections had a negative effect on pedestrian crash occurrences. Specifically, the presence of exclusive pedestrian signals for all crosswalks was found to significantly reduce the risk of pedestrian crashes by 43%. The present study is expected to shed more light on a deeper understanding of the environmental determinants of pedestrian crashes.

Rethinking safety in numbers: are intersections with more crossing pedestrians really safer?

Objective To advance the interpretation of the ‘safety in numbers’ effect by addressing the following three questions. How should the safety of pedestrians be measured, as the safety of individual pedestrians or as the overall safety of road facilities for pedestrians? Would intersections with large numbers of pedestrians exhibit a favourable safety performance? Would encouraging people to walk be a sound safety countermeasure? Methods We selected 288 signalised intersections with 1003 pedestrian crashes in Hong Kong from 2010 to 2012. We developed a Bayesian Poisson-lognormal model to calculate two common indicators related to pedestrian safety: the expected crash rate per million crossing pedestrians and the expected excess crash frequency. The ranking results of these two indicators for the selected intersections were compared. Results We confirmed a significant positive association between pedestrian volumes and pedestrian crashes, with an estimated coefficient of 0.21. Although people who crossed at intersections with higher pedestrian volumes experienced a relatively lower crash risk, these intersections may still have substantial potential for crash reduction. Conclusions Conclusions on the safety in numbers effect based on a cross-sectional analysis should be reached with great caution. The safety of individual pedestrians can be measured based on the crash risk, whereas the safety of road facilities for pedestrians should be determined by the environmental hazards of walking. Intersections prevalent of pedestrians do not always exhibit favourable safety performance. Relative to increasing the number of pedestrians, safety strategies should focus on reducing environmental hazards and removing barriers to walking.

Addressing the Modifiable Areal Unit Problem in Traffic Safety: Definition, Potential Solutions and Future Research

This study fully addressed the modifiable areal unit problem (i.e. MAUP) that was well-known in geography but generally ignored by safety analysts. The basic issue of MAUP was introduced firstly, and then four potential strategies, i.e. using disaggregate data as possible, capturing spatial non-stationarity, designing optimal zoning systems, conducting sensitivity analysis to report the scope and magnitude of MAUP, were proposed and illustrated to deal with the problem in an integrated way, followed by the future research directions suggested in the end. As model building and calibration activities are not independent of zone design, the MAUP should not only be addressed from the perspective of statistical analysis, but also need a geographical solution. Specifically, in traffic safety, more efforts are desired to calibrate the state-of-art modeling technique at various levels of data aggregation based on spatial homogeneity in traffic safety, transport characteristics, and

Developing an Optimized Artificial Neural Network to Predict Traffic Crash Injury Severity

A structure optimization algorithm for developing an artificial neural network (ANN) for predicting crash injury severity has been proposed in this study to improve its generalization capacity. Two-vehicle crash records in 2006 from the Florida Department of Highway Safety and Motor Vehicles (DHSMV) have been collected in order to demonstrate the proposed method and to compare it with a traditional statistical model -- an ordered logit (OL) model. According to the results, both the trained and optimized ANN models outperform the OL model in terms of predictive accuracy, which again displays its great potential for modeling crash injury severity. Although the accuracy of the training and testing dataset are equivalent in the trained and optimized ANN models in the case study, 33 of 210 connections in the trained ANN model have been deleted by the optimization algorithm. The computational burden of prediction has been reduced, which indicates the optimized ANN model to be a good alternative for crash injury severity analysis.