INFLUENCING FACTORS FOR RAIL-HIGHWAY GRADE CROSSING ACCIDENTS IN FLORIDA. FINAL REPORT

This report describes the estimation of a model for predicting rail-highway accident rates in Florida. Of 4,267 crossings with data from 1982 onwards, 4,231 were matched with their accident history for the years 1982, 1983, 1986, and 1987. Potential variables evaluated were trains per day, maximum train speed, number of tracks, crossbucks, stop signs, flashing lights, gates, number of parallel roads, required stopping sight distance, minimum clear quadrant stopping distance, number of lanes, posted highway speed, and average daily traffic. Stepwise logistic multiple regression was used for estimation. Accident rate is predicted as a function of exposure (essentially, 1 plus trains per day times average daily traffic), maximum train speed, posted highway speed, presence of stop signs, and presence of flashing lights.