A number of models have been applied to traffic injury severity analysis,among which is multinomial logit model. Compared to ordered discrete outcome models such as ordered logit model, multinomial logit model has the advantage of not restricting the effects of each variable on different injury outcomes to be the same. However, it is well known that multinomial logit model suffers from the independence from irrelevant alternatives (IIA)property. In this research, a latent class model is introduced for traffic injury severity analysis. Latent class model holds the advantages of the multinomial logit model. In addition, this new model can well solve the problems associated with the IIA property. Both the multinomial logit model and the latent class model are applied to analyze rural traffic injury severity data collected in Florida. Their model fitting, elasticities, andcross-elasticities are evaluated and compared.
Abstract