The Outlier Request

The request raises research integrity concerns. Removing 14 responses because they might weaken results is not the same as excluding data for a pre-specified, defensible reason. The pre-registered analysis plan does not mention completion-time exclusions, and the responses have complete answers and no failed attention checks. The participants were only slightly faster than the median, which is weak evidence of carelessness. Deleting them after seeing the data could bias results toward the supervisor's hypothesis. It also creates a transparency problem: the final analysis would not match the plan, and the paper or thesis might misrepresent what was done. There is a power differential: the supervisor controls my pay and reference letter, which makes it harder to object. I would also consider that the data file is shared and another assistant has already cleaned variables, so changes may be hard to trace. The concern is not that the supervisor is necessarily malicious; it may be a normalized questionable practice. But normal does not mean acceptable. I would want the original data preserved and any exclusion justified in writing.

I would respond calmly and ask for clarification rather than accuse. I might say, 'I want to make sure I understand the rationale. The pre-registration does not mention time-based exclusions. Are we planning to report this as a deviation, or is there a lab policy I am missing?' I would point out that the 14 responses passed the attention checks and were not dramatically faster. I would propose a less destructive alternative: keep all data and run a sensitivity analysis with and without the faster responses, then report both. That addresses the supervisor's worry about weak results without hiding data. I would ask whether we can document the decision and the reason. If the supervisor said 'everyone does this,' I would not argue about everyone; I would focus on what we can defend in the paper. I would also keep a copy of the original file so the data are not lost. I would try to stay collaborative, because the goal is good science, not winning a fight.

If the supervisor insisted, I would not delete the only copy of the data. I would preserve the original file and make a clear note of what I was asked to do. I would ask for the instruction in writing, not as a threat but as a way to document the analysis decision; a legitimate request should be defensible. I would then seek guidance from a trusted source: the lab's faculty advisor, the IRB, or a research integrity office, depending on the setting. I would frame it as a question about pre-registration and data handling, not as a personal attack. If the supervisor threatened my position or reference letter, I would recognize that as retaliation and keep records. I might also ask whether another senior lab member can review the analysis plan. I would not quietly comply, because deleting data to shape results is a serious breach that could harm participants, readers, and the lab. At the same time, I would avoid dramatic public accusations; the goal is to correct the process with the least unnecessary damage.