25

I submitted a paper to a journal, and the decision was "accept with minor revisions." As I was addressing the reviewer questions to submit a final copy, I realized that I may have misrepresented the data.

Specifically, I may have over-reached with a statistical model. The model involved repeated measures of individuals who were initially one group (let's call it condition A), but were later split into two groups (B and C). The model gives me differences between conditions. My mistake was to calculate the actual values at each condition based on these differences, which was wrong because by dumb luck the 2 groups had a systematic offset at A before they were split (even though the members of each group were chosen randomly). Hopefully that explanation makes some sense. I presented the data in terms of "actual" values rather than differences because I thought this would be more informative for the reader. So, basically, all the numbers are wrong. It's a stupid mistake in hindsight, since I could find this information in the raw data.

The conclusions of the paper and the statistically significant relationships would still be the same in the corrected version, because I believe the model to be the correct one. However, almost all of the numbers, tables, and figures would need modification. And of course the strange systematic offset between randomized groups will raise eyebrows.

I'm certainly going to notify the editor and attempt to submit a corrected version of the paper. I'm wondering what I should expect to happen, though. Are journals understanding of this type of error, or does it represent something so egregious that the paper would be rejected and my credibility would be damaged? Or would it simply be reviewed again? I understand that it's largely up to the editor to decide, but I'm wondering if anyone with similar experience might give me an idea of what I'm dealing with. Thanks!

  • What will happen next if the pictures are wrong? See here: retractionwatch.com/category/by-author/olivier-voinnet – Quora Feans Jul 11 '15 at 15:29
  • @QuoraFeans You seem to be comparing a case of willful misconduct to an accidental error. – March Ho Apr 26 '16 at 23:42
  • 1
    @MarchHo: once you realize the conclusions are wrong and do nothing, it is willful misconduct, not an accidental error anymore. – Quora Feans Apr 27 '16 at 11:48
26

Though I don't have any prior experience with this, I agree that you should notify the editor and should explain to them how and why this mess happened. Also let then know that it was a mistake and as soon as you realized it, you let them know. Also let them know that you are ready if they want to have the paper reviewed again. I believe that they would understand the mistake and would not make a scene out of it. But yes, you have to choose your words carefully and should politely explain them the point and admit your mistake.

8
  1. Put in the new statistics.
  2. Check whether the results are the same or deviate.
  3. If they deviate even slightly, the paper is a no go. Withdraw it. Don't try to make the new data reach the old results.
  4. If the results do not deviate, notify the editor of the error and changes, and ensure that you highlight that results do not change.
  • 1
    As the OP mentioned, the conclusions and the statistical relationships are still the same, but the numbers, tables, and figures have to be modified. – henning Jun 23 '15 at 9:16
  • The distinction you're drawing between the "statistics" and the "results" seems strange. Even if the sign and tests all point in the same direction, effects of a different magnitude might, or might not, be cause for concern or withdrawal. If you think the substantive takeaways from the paper are different, you should indeed withdraw the manuscript. If you think they are intact, you should highlight the differences so that the editors and reviewers can decide for themselves if they agree. I think this might be what you meant and I'm just confused by your reference to "the results." – Benjamin Mako Hill Apr 26 '16 at 23:19
6

It sounds like you screwed up the statistics. I would not want a publication with statistics that I know are wonky. I would withdraw the paper and appologize to the editor. I would ask the editor about resubmitting once the analysis issues are sorted. I would then bring in a statistical consultant to make sure your analysis is correct.

  • 32
    I'm not sure withdrawing is necessary: I think the decision of whether to accept the minor revision, re-review, or re-submit depends on the degree to which the final values and conclusions are actually affected, and should be left up to the editor. – jakebeal Jun 22 '15 at 22:08
  • @jakebeal To me withdrawing an unpublished paper is just not that big of a deal. Worse case is you have to submit to a new journal once the issues are fixed. The advantage of withdrawing and asking about resubmitting is you prevent the editor from having to make a difficult decision. – StrongBad Jun 23 '15 at 17:51
3

I have experienced a somewhat similar situation. In our case, it was the reviewer (who had given accept with minor changes), who wanted some clarification of a figure. When we started to clarify, we found out that a mistake made that figure, another figure and some numbers wrong.

What we did was to contact the editor and explained the situation, we cleared up the mess, and resubmitted with a letter to the reviewer explaining what had happened, and if the person would please look over the figures again. Both editor and reviewer were very understanding, and the article was accepted with the changes.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.