Abstract
Multiple regression was used to investigate the influence of physical factors on accident rates at 1,140 rail-highway grade crossings in Florida. Selection of the final model involved examination of residuals, transformation of variables. dummy variables and interaction terms. Variables in the final model were functions of average daily traffic, type of crossing protection, maximum train speed, trains per day. crossing speed limit, and number of lanes of traffic. Transformations were made to return estimates to the original scale of the data. Suggestions were made for improving the model and the data.