This study is to perform empirical analysis on observational before-after studies in Naive Method, Comparison Group(CG) Method and Empirical Bayes(EB) Method, and to compare with their results and to propose ways to apply to evaluation researches. For this purpose, the evaluation of road safety audit executed on Yŏng-dong freeway in 2005 and 2006 was performed. As a result, all three methods have showed improved effects due to safety treatments. The safety effectiveness of Naive method is the largest, CG Method is the second and EB method is the last. The results of Naive method are overestimated due to the trend of reducing traffic accidents and those of CG method are affected by the external casual effects of comparison group. In the EB method, as "regression to the mean" phenomenon are controlled by reference group's accident model, it's result is relatively more accurate than that of other methods. In the conduct of evaluation studies, the analysts have to understand the pros and cons of each evaluation method. And after leading the survey on accident trends of related all sites, evaluation analysis is performed to be able to minimize bias. Keywords comparison group method, empirical bayes method, naive method, observational before-after studies, regression to the mean, road safety audit 비교그룹에 의한 방법, 경험적 베이즈 방법, 단순비교법, 관찰적 사전․사후 평가연구, 평균으로의 회귀, 도로교통 안전진단.