Timeline for I have egregiously sloppy (possibly falsified) data that I need to correct. How should I go about this?
Current License: CC BY-SA 4.0
12 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Feb 25, 2021 at 17:36 | comment | added | 8263xiao | @PeterCordes So I did start fresh from the raw data and redid everything. I had changes I wanted to make to the analysis anyways because I learned a lot in the years after that project. It's just that the possibility that based on my past inaccuracies someone could reasonably conclude that I was acting dishonorably, even if I am not doing so now, even if I already informed my PI my past spreadsheets were flawed and their contents should not be the basis of the paper, that worries me. /fin | |
| Feb 25, 2021 at 17:25 | comment | added | 8263xiao | @PeterCordes I think the issue is, now looking back at it again, that the data inaccuracies are very suspicious and could be consistent with misconduct; if someone meticulously combs through my old spreadsheets, that is a conclusion they could easily draw as the inaccuracies made the graphs look better. It is true that the scientific record will not be tainted by this, and I was not stupid/malicious enough to tamper with raw data. 1/ | |
| Feb 25, 2021 at 15:26 | comment | added | Peter Cordes | @8263xiao: Since your bad data never got published, any mistakes (intentional or otherwise) are at least semi-private, and in your past. Clearly you know better now, and can just start fresh from the raw data (which you know hasn't been intentionally tampered with), with better analysis tools and methodology that you've learned since then. | |
| Feb 24, 2021 at 19:57 | comment | added | bob | I'm not sure honestly. Notes in the Excel workbook that indicate dishonest intent ("round down to minimize appearance of errors")? Maybe someone else could comment here? But basically if it doesn't scream "this looks like academic dishonesty", I'd take a deep breathe and relax. It probably wasn't. This of course is taking at face value an inability to remember what happened. | |
| Feb 24, 2021 at 19:19 | comment | added | 8263xiao | @bob What would be evidence to the contrary? | |
| Feb 24, 2021 at 18:38 | comment | added | bob | If you can't remember what happened, I'd give yourself the benefit of the doubt and assume it was an honest mistake unless and until you discover evidence to the contrary. Basically treat your work as you would someone else's. It's really no different at this point, since you can't remember what you did. Your work is as much a black box to you as someone else's would be. | |
| Feb 23, 2021 at 20:48 | comment | added | 8263xiao | I think my PI trusts me, perhaps undeservedly so back then as an undergrad. It is absolutely my responsibility to correct the data because it will be reflected in the scientific record and not doing so would tarnish us both. I just hope that he trusts that I'm now on top of things and that he's not painstakingly combing through what I already told him is invalid analysis, for both our sakes. /fin | |
| Feb 23, 2021 at 20:27 | comment | added | 8263xiao | My former PI never delved into the spreadsheets or compared them to raw data; I only showed him graphs. I don't think he ever pressured me to "improve" results per se, though as an undergrad I was definitely very insecure, eager to please, but not particularly professional/competent. I do want to get this published, but certainly not with bad data. I do have the original raw data that I now analyzed correctly, and I told him I found a number of errors in my past analysis. Hopefully he believes me and looks mainly at the updated/new analysis without delving into the past too much. /3 | |
| Feb 23, 2021 at 20:11 | comment | added | 8263xiao | It's possible that when I was analyzing the data I misremembered the setups for each trial and thought they were both the same, and I may have set up my spreadsheet that way and the mixups gave me the wrong impression that I had the additional points for both. But that does look suspicious. For my corrected version, I just reported fewer data points, consistent with what I saw in the correct raw data. It didn't really change the overall results but did make them a bit more noisy. 2/ | |
| Feb 23, 2021 at 20:00 | comment | added | 8263xiao | Now looking back again, the most suspicious thing I saw in the data was that I had copied only some of the results of different trial into a spreadsheet for a particular trial. While certainly that is possible with error (I may have had both trials' raw data open at the time and thus it was possible to mix up), there were more data points in the trial I erroneously copied from, which resulted in the original trial having more points than it actually did, matching the other trial. 1/ | |
| Feb 23, 2021 at 17:45 | review | First posts | |||
| Feb 23, 2021 at 20:11 | |||||
| Feb 23, 2021 at 17:45 | history | answered | emily | CC BY-SA 4.0 |