We aim to evaluate the safety effectiveness of already implemented countermeasures for wrong-way driving (which is pretty rare) at several locations. The before-after data is also available.
Dear Mahdi, it sounds like really interesting analysis. Before/after is of course typical design, but confounding factors (mostly general trend changes and regression-to-the-mean) have to be considered - otherwise results may be misleading (especially with low numbers which you mention). Therefore it is recommended to conduct before/after study with comparison group (to control for general trend) and empirical Bayes adjustment (to control for RTTM). More information you can find for example here - http://safety.fhwa.dot.gov/tools/crf/resources/fhwasa10032/fhwasa10032.pdf
Other possibility is to use some alternative (and more frequent) indicators such as traffic conflicts. However you would need some video record before the change in order to retrospectively detect conflicts.
Statistics are based on a rule of big numbers. The problem with accidents from statistical point of view is that:
- there are to scarse (sic!)
- there are many factors to be taken into account.
Therefore statistical methods usually do not work for single improvements. You can do statistics but without scientific validity or try with traffic conflicts.