Poisson, Poisson-gamma and zero-inflated regression models of motor vehicle crashes: balancing statistical fit and theory.

There has been considerable research conducted over the last 20 years focused on predicting motor vehicle crashes on transportation facilities. The range of statistical models commonly applied includes binomial, Poisson, Poisson-gamma (or negative binomial), zero-inflated Poisson and negative binomial models (ZIP and ZINB), and multinomial probability models. Given the range of possible modeling approaches and the host of assumptions with each modeling approach, making an intelligent choice for modeling motor vehicle crash data is difficult. There is little discussion in the literature comparing different statistical modeling approaches, identifying which statistical models are most appropriate for modeling crash data, and providing a strong justification from basic crash principles. In the recent literature, it has been suggested that the motor vehicle crash process can successfully be modeled by assuming a dual-state data-generating process, which implies that entities (e.g., intersections, road segments, pedestrian crossings, etc.) exist in one of two states-perfectly safe and unsafe. As a result, the ZIP and ZINB are two models that have been applied to account for the preponderance of "excess" zeros frequently observed in crash count data. The objective of this study is to provide defensible guidance on how to appropriate model crash data. We first examine the motor vehicle crash process using theoretical principles and a basic understanding of the crash process. It is shown that the fundamental crash process follows a Bernoulli trial with unequal probability of independent events, also known as Poisson trials. We examine the evolution of statistical models as they apply to the motor vehicle crash process, and indicate how well they statistically approximate the crash process. We also present the theory behind dual-state process count models, and note why they have become popular for modeling crash data. A simulation experiment is then conducted to demonstrate how crash data give rise to "excess" zeros frequently observed in crash data. It is shown that the Poisson and other mixed probabilistic structures are approximations assumed for modeling the motor vehicle crash process. Furthermore, it is demonstrated that under certain (fairly common) circumstances excess zeros are observed-and that these circumstances arise from low exposure and/or inappropriate selection of time/space scales and not an underlying dual state process. In conclusion, carefully selecting the time/space scales for analysis, including an improved set of explanatory variables and/or unobserved heterogeneity effects in count regression models, or applying small-area statistical methods (observations with low exposure) represent the most defensible modeling approaches for datasets with a preponderance of zeros.

[1]  D F Jarrett,et al.  Accidents at blackspots: Estimating the effectiveness of remedial treatment with special reference to the 'Regression-to-Mean' effect , 1981 .

[2]  John N. Ivan,et al.  Statistical challenges with modeling motor vehicle crashes: Understanding the implications of alternative approaches , 2004 .

[3]  Simon Washington,et al.  Empirical Investigation of Interactive Highway Safety Design Model Accident Prediction Algorithm: Rural Intersections , 2003 .

[4]  Guohua Pan,et al.  Local Regression and Likelihood , 1999, Technometrics.

[5]  W. Greene,et al.  Accounting for Excess Zeros and Sample Selection in Poisson and Negative Binomial Regression Models , 1994 .

[6]  Ron Goldman,et al.  Poisson approximation , 2000, Proceedings Geometric Modeling and Processing 2000. Theory and Applications.

[7]  M. Wand.  Local Regression and Likelihood , 2001 .

[8]  T. Louis,et al.  Bayes and Empirical Bayes Methods for Data Analysis. , 1997 .

[9]  Venkataraman N. Shankar,et al.  Modeling crashes involving pedestrians and motorized traffic , 2003 .

[10]  Bruce N. Janson,et al.  Diagnostic Methodology for the Detection of Safety Problems at Intersections , 2002 .

[11]  Fred Mannering,et al.  Impact of roadside features on the frequency and severity of run-off-roadway accidents: an empirical analysis. , 2002, Accident; analysis and prevention.

[12]  R Kulmala,et al.  SAFETY AT RURAL THREE- AND FOUR-ARM JUNCTIONS. DEVELOPMENT AND APPLICATION OF ACCIDENT PREDICTION MODELS. , 1995 .

[13]  Abraham P. Punnen,et al.  On the nature of the binomial distribution , 2001, Journal of Applied Probability.

[14]  Timothy C. Coburn,et al.  Statistical and Econometric Methods for Transportation Data Analysis , 2004, Technometrics.

[15]  Bradley P. Carlin,et al.  BAYES AND EMPIRICAL BAYES METHODS FOR DATA ANALYSIS , 1996, Stat. Comput..

[16]  Bhagwant Persaud,et al.  Accident Prediction Models With and Without Trend: Application of the Generalized Estimating Equations Procedure , 2000 .

[17]  J. Ivan,et al.  Explaining two-lane highway crash rates using land use and hourly exposure. , 2000, Accident; analysis and prevention.

[18]  Lianfen Qian,et al.  Nonparametric Curve Estimation: Methods, Theory, and Applications , 1999, Technometrics.

[19]  Nalini Ravishanker,et al.  Selecting exposure measures in crash rate prediction for two-lane highway segments. , 2004, Accident; analysis and prevention.

[20]  Bhagwant Persaud,et al.  ACCIDENT PREDICTION MODELS FOR FREEWAYS , 1993 .

[21]  K. Land,et al.  A Comparison of Poisson, Negative Binomial, and Semiparametric Mixed Poisson Regression Models , 1996 .

[22]  Simon Washington,et al.  Validation of FHWA Crash Models for Rural Intersections: Lessons Learned , 2003 .

[23]  Ingram Olkin,et al.  Probability Models and Applications , 2019 .

[24]  Dominique Lord,et al.  The prediction of accidents on digital networks, characteristics and issues related to the application of accident prediction models , 2000 .

[25]  Ezra Hauer,et al.  OBSERVATIONAL BEFORE-AFTER STUDIES IN ROAD SAFETY -- ESTIMATING THE EFFECT OF HIGHWAY AND TRAFFIC ENGINEERING MEASURES ON ROAD SAFETY , 1997 .

[26]  Stuart G. Baker,et al.  Multinomial–Poisson Transformation , 2006 .

[27]  Keizo Yoneda.  Estimations in some modified Poisson distributions. , 1962 .

[28]  J. G. Cragg.  Some Statistical Models for Limited Dependent Variables with Application to the Demand for Durable Goods , 1971 .

[29]  S. Miaou.  The relationship between truck accidents and geometric design of road sections: Poisson versus negative binomial regressions. , 1994, Accident; analysis and prevention.

[30]  Ezra Hauer,et al.  Estimation of safety at signalized intersections , 1988 .

[31]  Christian Gourieroux,et al.  A count data model with unobserved heterogeneity , 1997 .

[32]  F Mannering,et al.  Modeling accident frequencies as zero-altered probability processes: an empirical inquiry. , 1997, Accident; analysis and prevention.

[33]  P. Schmidt,et al.  Predicting Criminal Recidivism Using "Split Population" Survival Time Models , 1987 .

[34]  Hoong Chor Chin,et al.  Study of Intersection Accidents by Maneuver Type , 2002 .

[35]  J L Martin,et al.  Comparison of road crashes incidence and severity between some French counties. , 2003, Accident; analysis and prevention.

[36]  Q. Vuong.  Likelihood Ratio Tests for Model Selection and Non-Nested Hypotheses , 1989 .

[37]  Christopher Zorn,et al.  EVALUATING ZERO-INFLATED AND HURDLE POISSON SPECIFICATIONS , 1996 .

[38]  Samuel Kotz,et al.  Discrete Distributions: Distributions in Statistics , 1971 .

[39]  Fred L. Mannering,et al.  Negative binomial analysis of intersection accident frequencies , 1996 .

[40]  Zvi Drezner,et al.  A generalized binomial distribution , 1993 .

[41]  Diane Lambert,et al.  Zero-inflacted Poisson regression, with an application to defects in manufacturing , 1992 .

[42]  J. Mullahy.  Specification and testing of some modified count data models , 1986 .

[43]  Rajan Batta,et al.  Evaluating the reliability of automated collision notification systems. , 2003, Accident; analysis and prevention.

[44]  L. R. Taylor,et al.  Aggregation, Variance and the Mean , 1961, Nature.

[45]  H C Chin,et al.  Modeling Accident Occurrence at Signalized Tee Intersections with Special Emphasis on Excess Zeros , 2003, Traffic injury prevention.

[46]  W. Feller,et al.  An Introduction to Probability Theory and Its Application. , 1951 .

[47]  E. John Russell.  Rothamsted Experimental Station , 1944 .

[48]  Bradley P. Carlin,et al.  Generalized Linear Models for Small-Area Estimation , 1998 .

[49]  Siddhartha R. Dalal,et al.  Empirical bayes prediction for a compound poisson-multinomial process , 1990 .

[50]  David C. Heilbron,et al.  Zero-Altered and other Regression Models for Count Data with Added Zeros , 1994 .

[51]  Jye-Chyi Lu,et al.  Multivariate Zero-Inflated Poisson Models and Their Applications , 1999, Technometrics.

[52]  Shaw-Pin Miaou,et al.  Modeling Traffic Crash-Flow Relationships for Intersections: Dispersion Parameter, Functional Form, and Bayes Versus Empirical Bayes Methods , 2003 .

[53]  J. T. Wulu,et al.  Regression analysis of count data , 2002 .

[54]  Dominique Lord,et al.  Traffic Safety Diagnostic and Application of Countermeasures for Rural Roads in Burkina Faso , 2002 .

[55]  E Hauer,et al.  SAFETY ANALYSIS OF ROADWAY GEOMETRIC AND ANCILLARY FEATURES , 1996 .

[56]  K. Poortema,et al.  On modelling overdispersion of counts , 1999 .

[57]  Jerry Nedelman,et al.  Bernoulli Trials, Poisson Trials, Surprising Variances, and Jensen's Inequality , 1986 .

[58]  P. W. Wilson,et al.  Analyzing Frequencies of Several Types of Events: A Mixed Multinomial-Poisson Approach , 1990 .

[59]  E Hauer,et al.  Overdispersion in modelling accidents on road sections and in empirical bayes estimation. , 2001, Accident; analysis and prevention.