01477746

Component

This paper focuses on the relevance of alternate discrete choice frameworks for modeling driver injury severity. The study empirically compares the ordered response and unordered response models in the context of driver injury severity in traffic crashes. The alternative modeling approaches considered for the comparison exercise include: for the ordered response framework- ordered logit (OL), generalized ordered logit (GOL) and for the unordered response framework - multinomial logit (MNL), nested logit (NL) and ordered generalized extreme value logit (OGEV) model. A host of comparison metrics are computed to evaluate the performance of these alternative models. To our knowledge, the study provides a first of its kind comparison exercise of the performance of ordered and unordered response models for examining the impact of exogenous factors on the driver injury severity. The research also captures the effect of potential underreporting on alternative choice frameworks by artificially creating an underreported data sample from the driver injury severity sample. The empirical analysis is based on the 2010 General Estimates System (GES) data base. The comparison exercise clearly highlights the superiority of the GOL model on the estimation and the validation sample in terms of data fit compared to the OL and MNL models. The estimation with the artificial underreported sample consistently obtains the wrong elasticities and these errors are substantially reduced for both GOL and MNL models with the correction measures for the thresholds/constants of these models based on the true aggregate shares. The most striking finding is the fact that the MNL model does not perform any better in the underreporting context. In fact, the GOL elasticity effects of underreported estimates with corrections are closer to the true elasticity effects than that of the MNL model. Overall, the results of the empirical comparison provide credence to the belief that an ordered systems that allow for exogenous variable effects to vary across alternatives offer superior fit compared to unordered systems in modeling driver injury severity.

This paper was sponsored by TRB committee ABJ80 Statistical Methods.

01470560

13-4081

English

26p

2013

Transportation Research Board 92nd Annual Meeting

Digital/other

References; Tables

Choice models; Crash data; Injury severity; Logits; Multinomial logits; Variables

Underreporting (Traffic accident injuries)

Data and Information Technology; Highways; Safety and Human Factors; I71: Traffic Theory; I80: Accident Studies

Transportation Research Board Annual Meeting 2013 Paper #13-4081

TRIS, TRB, ATRI

Feb 5 2013 12:47PM