We created a dataset of about 15,500 stationary work zones including crash data.
We applied the Empirical Bayes (EB) before-and-after method in this study.
We estimated Crash Modification Factors for different work zone layouts.
We found that crossovers represent the critical work zone layouts for safety.