|
Abstract:
|
Influence of roadway and traffic factors on the likelihood of individual collision types have been studied through separate statistical models for each collision type. However, this approach can be very tedious and can produce unreliable estimate for rare collision types. To overcome these limitations, two-stage approaches have been proposed where a model which predicts total crash frequency is combined with proportion model to predict frequency of different collision types. Later, one-stage joint models, in which both models are estimated simultaneously, have also been proposed for macro-level analysis. This study investigates the performance of this joint model paradigm in analyzing crash frequency by collision type on individual road segments. For this, a joint negative binomial-multinomial fractional split (NB-MFS) model is used. Moreover, this study also proposes the use of a multinomial logit (MNL) model to estimate the proportion of different collision types. The goodness of fit statistics show that the NB-MNL model performs better than both NB-MFS and collision-specific NB models and is a promising approach in predicting crash frequency by collision type.
|
