This research employs the random parameters mixed logit (RPML) model to analyze pedestrian injury severity in pedestrian-vehicle crashes. RPML is compared with the behavioral models, such as the multinomial logit, heteroskedastic generalized extreme value, and error components mixed logit model. RPML is selected as the most appropriate model for pedestrian injury severity because of its superior fit and because this model accounts for heteroskedasticity and heterogeneity which this research finds to be useful inthe pedestrian context. The analysis was based on police-reported collision data from 1997 through 2000 from North Carolina. Several factors significantly increase the probability of fatal injury for pedestrians in pedestrian-vehicle crashes: pedestrian age, intoxicated drivers, darkness with or without streetlights, trucks, freeway, US route, state route, speeding involved crashes, and pedestrianÆs fault. Variables that were found to decrease the probability of fatal injury are: PM peak (15:00û17:59), walking along roadway, motorist turning or merging, and motorist backing up. Heteroskedasticity is induced by pedestrian age for fatal injury, and pedestrianheterogeneity is created by pedestrian gender for freeway and pedestrian fault and by pedestrian age for traffic sign and motorist backing up.
Samenvatting