Showing papers by "Dominique Lord published in 2016"

The Poisson inverse Gaussian (PIG) generalized linear regression model for analyzing motor vehicle crash data

Abstract: This article documents the application of the Poisson inverse Gaussian (PIG) regression model for modeling motor vehicle crash data The PIG distribution, which mixes the Poisson distribution and inverse Gaussian distribution, has the potential for modeling highly dispersed count data due to the flexibility of inverse Gaussian distribution The objectives of this article were to evaluate the application of PIG regression model for analyzing motor vehicle crash data and compare the results with negative binomial (NB) model, especially when varying dispersion parameter is introduced To accomplish these objectives, NB and PIG models were developed with fixed and varying dispersion parameters and compared using two data sets The results of this study show that PIG models perform better than the NB models in terms of goodness-of-fit statistics Moreover, the PIG model can perform as well as the NB model in capturing the variance of crash data Lastly, PIG models demonstrate almost the same prediction

Application of finite mixture models for analysing freeway incident clearance time

Abstract: A number of approaches have been developed for analysing incident clearance time data and investigating the effects of different explanatory variables on clearance time. Among these methods, hazard-based duration models (i.e. proportional hazard and accelerated failure time (AFT) models) have been extensively used. The finite mixture model is an alternative approach in survival data analysis, and offers greater flexibility in describing different shapes of the hazard function. Additionally, the finite mixture model assumes that the incident clearance time data set contains distinct subpopulations, and it allows the effects of explanatory variables to vary between different subpopulations. In this study, a g-component mixture model is applied to analyse incident clearance time. To demonstrate advantages of the proposed finite mixture model framework, incident clearance time data collected on freeway sections in Seattle, Washington State are analysed. Estimation and prediction results from the propo...

Sample-Size Guidelines for Recalibrating Crash Prediction Models: Recommendations for the Highway Safety Manual

Abstract: The Highway Safety Manual (HSM) prediction models are fitted and validated based on the crash data collected from a selected number of states in the United States. Therefore, for a jurisdiction to be able to fully benefit from applying these models, it is necessary to calibrate them to local conditions. The first edition of the HSM recommends calibrating the models using a one size fits-all sample-size of 30 to 50 locations with total of at least 100 crashes per year. However, the HSM recommendation is not fully supported by documented studies. The objectives of this paper are consequently to: 1) examine the required sample size based on the characteristics of the data that will be used for the recalibration process; and, 2) propose revised guidelines. The objectives were accomplished using simulation runs for different scenarios that characterized the sample mean and variance of the data. The simulation results indicate that as the ratio of the standard deviation to the mean (i.e., coefficient of variation) of the crash data increases, a larger sample-size is warranted to fulfil certain levels of accuracies. Taking this observation into account, sample-size guidelines were prepared based on the coefficient of variation of the crash data that are needed for the recalibration process. The guidelines were then successfully applied to the two observed datasets. The proposed guidelines can be used for all facility types and both for segment and intersection prediction models.

Improved Guidelines for Estimating the Highway Safety Manual Calibration Factors

Abstract: Crash prediction models can be used to predict the number of crashes and evaluate roadway safety. Part C of the first edition of the Highway Safety Manual (HSM) provides safety performance functions (SPFs). The HSM addendum that includes freeway and ramp chapters consist of severity distribution functions (SDFs) to estimate the crash severity as a function of geometric and traffic characteristics. In order to account for the differences in factors that were not considered or cannot be considered in the development of SPFs and SDFs, it is essential to calibrate them when they are applied to a new jurisdiction. The HSM recommends a one‐size‐fits-all sample size for calibration procedures that require crash data collected from randomly selected sites. However, the recommended sample size is not fully supported by documented studies, and several agencies have initiated SPF calibration efforts. In addition, there are no clear guidelines on when an agency should update their calibration factors (C-factors) and how they should make a decision on the need of region-specific calibration factors. The objectives of this research are to (1) review and document issues with the existing calibrating method in the HSM, (2) identify factors that influence the selection of the sample size for the SPFs calibration (or recalibration), (3) determine how frequently or when an agency should update their calibration factors, (4) determine whether or not having region-specific C-factors are justified and when they are needed, and (5) identify factors that influence the selection of the sample size for the SDFs calibration (or recalibration). The study objectives were accomplished using simulated and observed data. The guidelines included a discussion on (1) the sample size that is required to calibrate SPFs; (2) when the models should be recalibrated; (3) when the region-specific C-factors are recommended; and (4) the sample size that is required to calibrate SDFs.

A Flexible Modeling Approach Using Dirichlet Process Mixtures: Application to Multi-Level Railway Grade Crossing Crash Data

Abstract: This paper introduces a new approach to addressing two of the most challenging issues in road safety research, namely, how to account for unobserved heterogeneity and how to identify latent subpopulations in data. Compared to the approaches of applying random effects/parameters models and finite mixtures, the proposed approach employs a Bayesian semi-parametric methodology based on Dirichlet process mixtures. Our method has four noteworthy advantages: (i) it allows examining the robustness of distributional assumptions in random effects/parameters models; (ii) it allows identifying latent clusters in data; (iii) it enables identification of outliers (extreme observations) while allowing accommodating them in analyses without compromising the quality of estimates; and (iv) it is capable of estimating the number of latent clusters in data using an elegant mathematical structure. In this paper, we evaluate the proposed method on a railway grade crossing crash dataset with hierarchical (multilevel) structure, at municipality level, from Canada for the years 2008 to 2013. We use cross-validation predictive densities and pseudo Bayes factor for Bayesian model selection. While confirming the need for the multilevel modeling approach, the results pointed out the inadequacy of the parametric assumption. In fact, our proposed method improved model fitting significantly for the municipality-level data. In a fully probabilistic framework, we also identified the expected number of latent clusters with similar unknown/unmeasured features among 81 Canadian municipalities. It is possible thus to further investigate the reasons behind such similarities and dissimilarities, which could have important policy implications in terms of safety management process.