The Art of Regression Modeling in Road Safety

Hauer, Ezra

doi:10.1007/978-3-319-12529-9

Cited by 111 publications

(189 citation statements)

References 3 publications

Supporting

Mentioning

184

Contrasting

Unclassified

Order By: Relevance

“…The Poisson distribution has often been used to model the probability that an event occurs given a certain incident rate, λ. Several studies determined that the Poisson distribution can be used to approximate count incident generation when the incident rate is small (32)(33)(34)(35). However, the observed total runway incursions for this study showed that the sample mean was 3.8 and the variance 18.7 (variance greater than the mean).…”

Section: Model Developmentmentioning

confidence: 62%

“…Instead, the negative binomial distribution is used. Two fundamental assumptions are made in modeling with the negative binomial distribution: (a) runway incursions are Poisson distributed and (b) the populations of means are gamma distributed (35). However, for longitudinally clustered count data, such as the number of runway incursions from a set of airports, the observations from the same airports may not be mutually independent because some of the measured airport traits, such as the number of operations or weather, fluctuate over time.…”

Section: Model Developmentmentioning

confidence: 99%

See 1 more Smart Citation

Enhancing Safety Risk Management with Quantitative Measures

Claros

Sun

Edara

2017

Transportation Research Record: Journal of the Transportation R

View full text Add to dashboard Cite

At the airfield in hub airports, many activities occur that involve a range of participants, including various-size aircraft, ground vehicles, and workers. The safety management system is FAA's approach for systematically managing aviation safety. A major component of the safety management system is safety risk management (SRM), which entails analysis, assessment, and control of safety risks, including risks on the airfield. Current SRM has few specific safety models to estimate the likelihood or frequency of risks. This paper presents an example for development and incorporation of safety models into SRM. Specifically, it discusses safety models for runway incursion that use the following variables: total and general aviation operations, length of runway by type, number of taxiway intersections, snowfall, precipitation, number of hot spots, and construction activity. Categorization and processing of data were significant because each variable used could take on multiple forms, and some types of data involved review of airfield diagrams. The data used were from 137 U.S. hub airports for 2009 through 2014. For modeling, the negative multinomial distribution was used because it proved suitable for representing overdispersed data such as runway incursion frequency. Performance of the models was assessed through the goodness-of-fit measures of log likelihood, overdispersion, and cumulative residual plots. Models were developed for five severity categories of runway incursions and three types of surface events. The safety modeling approach presented here can serve as a foundation for development of other safety models that can be integrated into SRM to enable quantitative analysis of safety risks.The safety management system (SMS) is a formalized and proactive approach to system safety that directly supports the FAA's mission to provide the safest and most efficient aerospace system. SMS is an integrated set of processes consisting of four components: safety policy, safety assurance, safety promotion, and safety risk management (SRM). SRM describes the system; identifies the hazards; and analyzes, assesses, and controls risk (1). SRM requires specific quantitative measures to determine risk likelihood. For instance, the expected number of runway incursions in a system is a crucial measure for runway safety analysis. This paper proposes the enhancement of SRM with quantitative models of the frequency of runway incursions as a function of airport geometry, operations, construction activity, and weather variables. The most important components of SRM are summarized to illustrate the application of the models. In addition, a review of current aviation safety modeling is presented. Description of srMSRM safety analysis is divided into five phases that apply to activities such as operations, maintenance, procedures, or equipment development: Phase 1, describing the system; Phase 2, identifying hazards; Phase 3, analyzing risk; Phase 4, assessing risk; and Phase 5, treating risk. phase 1. Describing the system Describing th...

show abstract

Section: Model Developmentmentioning

confidence: 62%

Section: Model Developmentmentioning

confidence: 99%

Enhancing Safety Risk Management with Quantitative Measures

Claros

Sun

Edara

2017

Transportation Research Record: Journal of the Transportation R

View full text Add to dashboard Cite

show abstract

“…The Poisson distribution has been used in accident modeling since the work of L. von Bortkiewitcz who, in 1989, undertook analyses of the number of deaths by horse-kicks over a period of 20 years (Hauer, 2015). Because of its practicality, research has commonly used this distribution to estimate the SPF (Bezerra, Silva, Bastos, & Ferraz, 2011;Caliendo, Guida, & Parisi, 2007;Jovanis & Chang, 1986;Miao, 1993).…”

Section: Crash Prediction Analysismentioning

confidence: 99%

“…Although it represents a good starting point to fit safety models, the simple application of the Poisson distribution has had some inconveniencies on real transportation systems due to the nature of the crash generation process (Hauer, 2015;Lord & Mannering, 2010). The Negative Binomial regression is largely used (Ackaah & Salifu, 2011;Cafiso, Di Graziano, Di Silvestro, La Cava, & Persaud, 2010;Harwood et al, 2000;Lord & Persaud, 2000;McCullagh & Nelder, 1989;Srinivasan & Bauer, 2013) because of the possibility of accounting for the overdispersion of crash frequency.…”

Section: Crash Prediction Modelsmentioning

confidence: 99%