Examining driver behavior at the onset of yellow in a traffic simulator environment: Comparisons between random parameters and latent class logit models.

This study involves an examination of driver behavior at the onset of a yellow signal indication. Behavioral data were obtained from a driving simulator study that was conducted through the National Advanced Driving Simulator (NADS) laboratory at the University of Iowa. These data were drawn from a series of events during which study participants drove through a series of intersections where the traffic signals changed from the green to yellow phase. The resulting dataset provides potential insights into how driver behavior is affected by distracted driving through an experimental design that alternated handheld, headset, and hands-free cell phone use with "normal" baseline driving events. The results of the study show that male drivers ages 18-45 were more likely to stop. Participants were also more likely to stop as they became more familiar with the simulator environment. Cell phone use was found to some influence on driver behavior in this setting, though the effects varied significantly across individuals. The study also demonstrates two methodological approaches for dealing with unobserved heterogeneity across drivers. These include random parameters and latent class logit models, each of which analyze the data as a panel. The results show each method to provide significantly better fit than a pooled, fixed parameter model. Differences in terms of the context of these two approaches are discussed, providing important insights as to the differences between these modeling frameworks.