I have tried to do a regression discontinuity design on my data for my dissertation. I am trying to find the effectiveness of post-secondary remediation in North Dakota. One item that is giving me issues is that in North Dakota students are able to retest, thus reducing the validity of an RDD. The Math Cut off score for remediation is a 22 on the ACT. Those at or above a 22 will not be placed in remediation and those below the 22 are supposed to be placed in remediation. However, This creates testing manipulation because those who do not receive a 22 can retest until they are above that threshold. This leads to selection bias because not all of them will choose to retake the exam. In my data, there is a large difference in observable data points at the 21 and 22 scores. Are there any other analysis designs that I could utilize for this study? Any help would be greatly appreciated.

Thank you,
Russ Ziegler