Using graphical Markov models and Gibbs sampling to reconstruct vehicle/pedestrian accidents.

This paper illustrates how a statistical simulation technique called Gibbs Sampling can be applied to data commonly collected in the investigation of vehicle/pedestrian accidents, producing Bayes estimates of the vehicle's initial speed and location at the point of perception. A general discussion of how Bayesian reasoning applies to accident reconstruction is first presented, followed by a description of the collision model. The approach is then applied to the reconstruction of two accidents. (A)