Validating Ramp Crash Prediction Models Used for Roadway Design Decision-Making with a Negative Binomial Generalized Linear Model

• Activity Number: 496
• Type: Contributed
• Date/Time: Wednesday, August 12, 2015 : 8:30 AM to 10:20 AM
• Sponsor: Transportation Statistics Interest Group
• Author(s): Lindsay M. Lucas* and Karin M. Bauer
• Companies: MRIGlobal and MRIGlobal
• Keywords: Model Validation ; Crash Prediction Model ; Negative Binomial Model ; Decision-Making ; AASHTO Highway Safety Manual

Abstract

The American Association of State Highway and Transportation Officials (AASHTO) provides roadway guides and standards to highway and transportation departments. The Highway Safety Manual (HSM) is the guide used most by traffic safety engineers; it equips practitioners with quantitative tools to consider safety when making decisions related to design and operation of roadways. The HSM provides a method to predict average crash frequencies on several types of ramps by incorporating curve features into the prediction model. Key questions answered in this project were (1) to what extent did the HSM ramp model accurately predict crashes at loop and diamond ramps and (2) is the inclusion of curve features into a base model sufficient to differentiate between various ramp configurations? We used ramp crash data from California and Washington and predicted crash counts for each ramp using the HSM method. To compare predicted and observed crash frequencies per exposure, we used a generalized linear model with a negative binomial distribution, a logit link, and a predicted/observed indicator variable. The ratios of predicted to observed crash rates were estimated and statistically compared.