Paul P Jovanis

Analysis of Road Crash Frequency with Spatial Models

Despite the evident spatial character of road crashes, limited research has been conducted in road safety analysis to account for spatial correlation; further, the practical consequences of this omission are largely unknown. The purpose of this research is to explore the effect of spatial correlation in models of road crash frequency at the segment level. Different segment neighboring structures are tested to establish the most appropriate one in the context of modeling crash frequency in road networks. A full Bayes hierarchical approach is used with conditional autoregressive effects for the spatial correlation terms. Analysis of crash, traffic, and roadway inventory data from a rural county in Pennsylvania indicates the importance of including spatial correlation in road crash models. The models with spatial correlation show significantly better fit to the data than the Poisson lognormal model with only heterogeneity. Parameters significantly different from zero included annual average daily traffic (AADT) and shoulder widths less than 4 ft and between 6 and 10 ft. In four models with spatial correlation, goodness of fit was improved compared with the model including only heterogeneity. More important yet is the potential of spatial correlation to reduce the bias associated with model misspecification, as shown by the change in the estimate of the AADT coefficient and other parameters.

DETERMINATION OF NUMBER OF PROBE VEHICLES REQUIRED FOR RELIABLE TRAVEL TIME MEASUREMENT IN URBAN NETWORK

Several intelligent vehicle—highway system demonstration projects are currently assessing the feasibility of using probe vehicles to collect realtime traffic data for advanced traffic management and information systems. They have used a variety of criteria to determine the number of probes necessary, but few generalizable algorithms have been developed and tested. The described algorithm explicitly considers the time period for travel time estimation (e.g., 5, 10, or 15 min), the number of replications of travel time desired for each link during each measurement period (reliability criterion), the proportion of links to be covered, and the length of the peak period. This algorithm is implemented by using a simulation of the Sacramento Network (170 mi2) for the morning peak period. The results indicate that the number of probe vehicles required increases nonlinearly as the reliability criterion is made more stringent. More probes are required for shorter measurement periods. As the desired proportion of li...

Bayesian Multivariate Poisson Lognormal Models for Crash Severity Modeling and Site Ranking

Traditionally, highway safety analyses have used univariate Poisson or negative binomial distributions to model crash counts for different levels of crash severity. Because unobservables or omitted variables are shared across severity levels, however, crash counts are multivariate in nature. This research uses full Bayes multivariate Poisson lognormal models to estimate the expected crash frequency for different levels of crash severity and then compares those estimates to independent or univariate Poisson lognormal estimates. The multivariate Poisson log-normal model fits better than the univariate model and improves the precision in crash-frequency estimates. The covariances and correlations among crash severities are high (correlations range from 0.47 to 0.97), with the highest values found between contiguous severity levels. Considering this correlation between severity levels improves the precision of the expected number of crashes. The multivariate estimates are used with cost data from the Pennsylvania Department of Transportation to develop the expected crash cost (and excess expected cost) per segment, which is then used to rank sites for safety improvements. The multivariate-based top-ranked segments are found to have consistently higher costs and excess costs than the univariate estimates, which is due to higher multivariate estimates of fatalities and major injuries (due to the random effects parameter). These higher estimated frequencies, in turn, produce different rankings for the multivariate and independent models. The finding of a high correlation between contiguous severity levels is consistent with some of the literature, but additional tests of multivariate models are recommended. The improved precision has important implications for the identification of sites with promise (SWiPs), because one formulation includes the standard deviation of crash frequencies for similar sites as part of the assessment of SWiPs.

Disaggregate model of highway accident occurrence using survival theory

The analysis of discrete accident data and aggregate exposure data frequently necessitates compromises that can obscure the relationship between accident occurrence and potential causal risk components. One way to overcome these difficulties is to develop a model of accident occurrence that includes accident and exposure data at a mathematically consistent disaggregate level. This paper describes the conceptual and mathematical development of such a model using principals of survival theory. The model predicts the probability of being involved in an accident at time t given that a vehicle has survived until that time. Several alternative functional forms are discussed including additive, proportional hazards and accelerated failure time models. Model estimation is discussed for the case in which both accident and nonaccident trips are included and for the case with only accident data. As formulated, the model has the distinct advantage of being able to consider accident and exposure data at a disaggregate level in an entirely consistent analytic framework. A conditional accident analysis is undertaken using truck accident data obtained from a major national carrier in the United States. Model results are interpretable and generally reasonable. Of particular interest is that segmenting accidents in several categories yields very different sets of significant parameters. Driver service hours seemed to most strongly effect accident risk: regularly scheduled drivers who take frequent trips are likely to have a reduced risk of an accident, particularly if they have a longer (greater than eight) number of hours off-duty just prior to a trip.

Multiday driving patterns and motor carrier accident risk: A disaggregate analysis

A method has been developed to estimate the relative accident risk posed by different patterns of driving over a multiday period. The procedure explicitly considers whether a driver is on duty or off duty for each half hour of each day during the period of analysis. From a data set of over 1,000 drivers, nine distinct driving patterns are identified. Membership in the patterns is determined exclusively by the pattern of duty hours for seven consecutive days; for some drivers an accident occurred on the eighth day while others had no accident, therefore each pattern can be associated with a relative accident risk. Additional statistical modeling allowed the consideration, in addition to driving pattern, of driver age, experience with the firm, hours off duty prior to the last trip and hours driving on the last trip (either until the accident or successful completion of the trip). The finding of the modeling is that driving patterns over the previous seven days significantly affect accident risk on the eighth day. In general, driving during the early and late morning (e.g., midnight to 10 A.M.) has the highest accident risk while all seven other multiday patterns had indistinguishable risk. Consecutive hours driven also has a significant effect on accident risk: the first hour through the fourth hour having the lowest risk with a fluctuating increase in risk to a maximum beyond nine hours. Driver age and hours off duty immediately prior to a trip do not appear to affect accident risk significantly. These findings quantitatively assess the relative accident risk of multiday driving patterns using data from actual truck operations. Further research is recommended in the areas of refining model structures, adding explanatory variables (such as highway type), and testing more complex models.

Spatial Correlation in Multilevel Crash Frequency Models: Effects of Different Neighboring Structures

Recent research has shown the importance of spatial correlation in road crash models. Because many different spatial correlation structures are possible, however, this study tested several segment neighboring structures to establish the most promising one to model crash frequency in road networks. A multilevel approach was also used to account for the spatial correlation between road segments of different functional types, which are usually analyzed separately. The study employed a full Bayes hierarchical approach with conditional autoregressive effects for the spatial correlation terms. Analyses of crash, traffic, and roadway inventory data from rural engineering districts in Pennsylvania and Washington affirmed the importance of spatial correlation in road crash models. Pure distance-based neighboring models (i.e., exponential decay) performed poorly compared with adjacency-based or distance order models. The results also suggest that spatial correlation is more important in distances of 1 mi or less. The inclusion of spatially correlated random effects significantly improved the precision of the estimates of the expected crash frequency for all segments by pooling strength from their neighbors and thus reducing their standard deviation. Results from Pennsylvania and Washington showed that spatial correlation substantially increased the random effects. There was a consistent indication that 70% to 90% of the variation explained by the random effects resulted from spatial correlation. This suggests that spatial models offer a significant advantage, since poor estimates that result from small sample sizes and low sample means are a frequent issue in highway safety analysis. Application of spatial correlation to the identification of sites with promise indicated that more sites were identified because of a reduction in the variance of the estimates, which would allow for greater confidence in the selection of sites for treatment.

Crashes and crash-surrogate events: Exploratory modeling with naturalistic driving data

There is a need to extend and refine the use of crash surrogates to enhance safety analyses. This is particularly true given opportunities for data collection presented by naturalistic driving studies. This paper connects the original research on traffic conflicts to the contemporary literature concerning crash surrogates using the crash-to-surrogate ratio, π. A conceptual structure is developed in which the ratio can be estimated using either a Logit or Probit formulation which captures context and event variables as predictors in the model specification. This allows the expansion of the crash-to-surrogate concept beyond traffic conflicts to many contexts and crash types. The structure is tested using naturalistic driving data from a study conducted in the United States (Dingus et al., 2005). While the sample size is limited (13 crashes and 38 near crashes), there is reasonable correspondence between predicted and observed crash frequencies using a Logit model formulation. The paper concludes with a summary of empirical results and suggestions for future research.

Method for Identifying Factors Contributing to Driver-Injury Severity in Traffic Crashes

Numerous driver, vehicle, roadway, and environmental factors contribute to crash-injury severity. In addition to main effects, interactions between factors are very likely to be significant. The large number of potentially important factors, combined with the complex nature of crash etiology and injury outcome, present significant challenges to the safety analyst, who must select from a large number of factors and specify a comprehensive but feasible set of main factors and interactions for testing in statistical models. In addition, some factors contain a relatively large number of categories (e.g., weather conditions), and the selection of cut-off points for categorization of continuous factors may not be readily obvious (e.g., driver age). It is also important that statistical tests underlying these analyses accurately address the frequent problem of data sparseness. The development and testing of a variable-selection procedure to address each of these problems is the stated objective. Bus-involved crash data for Freeway 1 in Taiwan from 1985 through 1993 were used to screen a set of 39 possible influential factors, along with interactions. The final log-linear model shows that late-night or early-morning driving increases the risk for bus drivers of being severely injured, particularly when the drivers caused the accident or when the drivers were involved in rear-end accidents. Bus accidents involving large trucks or tractor-trailers also increase the risk. An assessment of the importance of considering interactions in crash models is presented as a conclusion.

INVESTIGATING EFFECT OF ADVANCED TRAVELER INFORMATION ON COMMUTER TENDENCY TO USE TRANSIT

A study of the effects of advanced transit information on commuter propensity to use transit is described. A computer-aided telephone interview was designed and conducted in Sacramento and San Jose, California. The most important information items that nontransit users seek about the transit services available to them is investigated. Using a customized stated preference choice sets, the likelihood that the commuters will use transit is explored. Commuter perceptions of transit and transit user satisfaction with the information currently available to them are studied. The results indicated that about 38 percent of the respondents who currently do not use transit might consider transit if the appropriate information is available. Analyses using the stated-preference technique and targeting this group of respondents indicated a promising effect of transit information systems in encouraging transit ridership if the desired information is provided. Using binary logit and ordered probit modeling produced resul...

Safety Effectiveness of Lane and Shoulder Width Combinations on Rural, Two-Lane, Undivided Roads

There is a need to evaluate low-cost safety strategies that states may implement as part of their Strategic Highway Safety Plan. FHWA organized a pooled fund study of 26 states to evaluate several low-cost safety strategies, including the reallocation of total paved width. This study identifies whether it is safer to increase lane width or increase shoulder width given a fixed total width. Geometric, traffic, and crash data were obtained for more than 52,000 mi of roadway in Pennsylvania and Washington State. A case-control approach was applied to evaluate the safety effectiveness of various lane–shoulder configurations. There was a general reduction in the odds ratio as total paved width, lane width, and shoulder width increased individually; this is consistent with previous research. However, the primary research objective was to estimate the safety effectiveness of reallocating a fixed total paved width. Individual state results did not indicate a clear trade-off between lane and shoulder width for a fixed total width. Supplementing the results of this study with previous research, crash modification factors (CMFs) are provided for several lane–shoulder combinations. The selected values present a more apparent trade-off, indicating a slight benefit to increasing lane width for a fixed total width. Importantly, the results differ from other studies that developed CMFs without considering the interaction between lane and shoulder width, including those studies currently referenced in the Highway Safety Manual. This raises the question of whether CMFs should reflect the interaction between lane and shoulder width.