Analyzing Different Parameterizations of the Varying Dispersion Parameter as a Function of Segment Length
Authors:
- 1 January 2009
- journal article
- research article
- Published by SAGE Publications in Transportation Research Record: Journal of the Transportation Research Board
- Vol. 2103 (1) , 108-118
Abstract
Until a few years ago, the dispersion parameter of Poisson–gamma models had been assumed to be invariant of the characteristics of the observations under study, but recent research in highway safety has shown that the dispersion parameter can depend on the covariates of the model. To account for this dependence, some researchers have reported that the dispersion parameter should be modeled solely as a function of segment length. The primary objective of this research was to examine empirically whether the dispersion parameter should be characterized using only the length of the segment. If not, the secondary objective consisted of determining alternative parameterizations using other covariates that would offer a better approach for characterizing the variance function of Poisson–gamma models. To accomplish the study objectives, 10 parameterizations describing the varying dispersion parameter were estimated with three different data sets collected in Texas, California, and Washington State. Flow-only models were used for comparing the parameterizations. The Akaike information criterion and other related goodness-of-fit (GOF) measures were used for evaluating and comparing the different models. The results of this study show that no single functional form or parameterization is suitable for all the data sets. Traffic flow was more significantly associated with the structured variation observed in the data than segment length. It is therefore recommended that transportation safety analysts evaluate different parameterizations and select the most appropriate one using a combination of GOF criteria, including the significance of the model's coefficients.Keywords
This publication has 21 references indexed in Scilit:
- Application of the Conway–Maxwell–Poisson generalized linear model for analyzing motor vehicle crashesAccident Analysis & Prevention, 2008
- Adjustment for Maximum Likelihood Estimate of Negative Binomial Dispersion ParameterTransportation Research Record: Journal of the Transportation Research Board, 2008
- Effects of Varying Dispersion Parameter of Poisson–Gamma Models on Estimation of Confidence Intervals of Crash Prediction ModelsTransportation Research Record: Journal of the Transportation Research Board, 2008
- On the nature of over-dispersion in motor vehicle crash prediction modelsAccident Analysis & Prevention, 2007
- Accident prediction model for railway-highway interfacesAccident Analysis & Prevention, 2006
- Poisson, Poisson-gamma and zero-inflated regression models of motor vehicle crashes: balancing statistical fit and theoryAccident Analysis & Prevention, 2005
- Modeling Traffic Crash-Flow Relationships for Intersections: Dispersion Parameter, Functional Form, and Bayes Versus Empirical Bayes MethodsTransportation Research Record: Journal of the Transportation Research Board, 2003
- Overdispersion in modelling accidents on road sections and in Empirical Bayes estimationAccident Analysis & Prevention, 2001
- Identification of sites for road accident remedial work by Bayesian statistical methods: an example of uncertain inferenceAdvances in Engineering Software, 2001
- Negative Binomial Analysis of Intersection-Accident FrequenciesJournal of Transportation Engineering, 1996