Vehicle headways play an important role in the analysis of highway capacity and level of service. This paper presents statistical procedures to identify statistical models for headway distributions, estimate their parameters, and test the goodness of fit. A four-stage identification process is suggested, which compares graphically empirical and theoretical density functions and hazard functions, and calculates two descriptive measures using the first four moments. In parameter estimation, modified methods are proposed in order to avoid the problems of the standard estimation methods. The superiority of parametric Monte Carlo methods over standard nonparametric methods in goodness-of-fit tests is demonstrated. It is also shown that the power of the tests can be further enhanced by calculating combined significance probabilities. Results are presented for the exponential, gamma, lognormal and semi-Poisson distributions. The data are from Finnish two-lane roadways, but the methods can be applied to any headway data. (A)
Abstract