Abstract
Errors and inconsistencies in crash location records are realistic problems that often compromise accuracy of safety model outcomes. This paper proposes a new safety analysis framework where an optimal network design module that estimates the most-probable site for each crash is combined with the standard regression analysis. An effective solution algorithm based on Lagrangian relaxation is developed for the network design model, and an iterative computation approach is proposed for location estimation and statistical regression. The proposed models are implemented with empirical dataand numerical results are presented. (A) Reprinted with permission from Elsevier.