Evaluating alternate discrete outcome frameworks for modeling crash injury severity.

摘要: This paper focuses on the relevance of alternate discrete outcome frameworks for modeling driver injury severity. The study empirically compares ordered response and unordered models in context severity traffic crashes. alternative approaches considered comparison exercise include: framework-ordered logit (OL), generalized (GOL), mixed (MGOL) framework-multinomial (MNL), nested (NL), extreme value (OGEV) multinomial (MMNL) model. A host metrics are computed to evaluate performance these models. provides a comprehensive examining impact exogenous factors research also explores effect potential underreporting by artificially creating an underreported data sample from sample. empirical analysis is based 2010 General Estimates System (GES) base—a nationally representative road crashes collected compiled about 60 jurisdictions across United States. examined model estimation validation (at aggregate disaggregate level). Further, presence explored, with without corrections estimates. results extensive analyses point toward emergence GOL framework as strong competitor MMNL