10
$\begingroup$

If a question A is a duplicate of a combination of two questions, what should I do? Should I flag it as duplicate of one of the two questions, and leave a comment?

The issue I have with comments is that that aren't meant to be kept forever: they may be deleted anytime, and users don't have access to their deleted comments. Even the source of non-deleted comments cannot be obtained.


Example 1: Neural networks: Differences between standard backpropagation, momentum and weight decay is a duplicate of the combination of Difference between neural net weight decay and learning rate and What is the intuition of momentum term in the neural network back propagation?.

Example 2: fine-tuned end-to-end meaning = Fine Tuning vs Joint Training vs Feature Extraction + What does end to end mean in deep learning methods?

$\endgroup$
  • 2
    $\begingroup$ I think the issue is real (albeit rate), but your example is not very good. Question A in your example is so bad that I think it should be downvoted for lack of research effort & closed as too broad. $\endgroup$ – amoeba says Reinstate Monica Jan 29 '17 at 21:10
  • 2
    $\begingroup$ Other than that, Should I flag it as duplicate of one of the two questions, and leave a comment? - that is what I usually do, yes. In fact, it is technically possible to close a question as a duplicate of two questions; I have seen it multiple times (but don't think I can find an example right now). I believe this can only be done via moderator intervention, perhaps one of the mods will comment to clarify. $\endgroup$ – amoeba says Reinstate Monica Jan 29 '17 at 21:12
  • $\begingroup$ If the situation is that the current question is really two questions, then one reasonable option is link the two duplicates in comments and vote to close as too broad. $\endgroup$ – Glen_b -Reinstate Monica Jan 30 '17 at 0:30
  • $\begingroup$ @Glen_b even if one doesn't think the question is too broad? $\endgroup$ – Franck Dernoncourt Jan 30 '17 at 0:33
  • 1
    $\begingroup$ Example: stats.stackexchange.com/questions/258820 - this got closed as a dup of two questions, and actually without any mod intervention. $\endgroup$ – amoeba says Reinstate Monica Jan 30 '17 at 0:44
  • 2
    $\begingroup$ @FranckDernoncourt If there is a post with two essentially separate questions that can stand alone (even if they're related), they should normally be asked as two questions each in their own post. If it isn't really two questions, it should probably stand on its own. We tend to often let lists of questions stay if there aren't so many of them (and just answer separate questions in one answer), but the principle is there and could be used as a way of dealing with this circumstance. $\endgroup$ – Glen_b -Reinstate Monica Jan 30 '17 at 1:00
7
$\begingroup$

A question can be closed as a duplicate of two others if different users voting to close pick different duplicates (a recent example), but a single user can't pick more than one. Leaving a comment linking to a second duplicate is good enough—it could be deleted, but won't be. Links to both will appear 'Linked' list on the right of the page.

In some cases, especially when A isn't asked as "B? and C?", a "stub" answer may be a good idea: one that disentangles the issues & explains how the linked duplicates address them. In others there may be some mileage in a "compare & contrast"-type answer.

$\endgroup$
  • $\begingroup$ (+1) Out of curiosity, can mods add another duplicate link to a question that is already closed as a duplicate? $\endgroup$ – amoeba says Reinstate Monica Jan 31 '17 at 21:54
  • 1
    $\begingroup$ @Amoeba In most cases, there is no apparent way to do that. $\endgroup$ – whuber Jan 31 '17 at 22:29
1
$\begingroup$

If the new question is distinct enough that you need to combine two old questions to made it a pseudo-duplicate, then it's not a duplicate. Leave it open and, if you want, use the information from the other two posts to create an answer to the combined question.

$\endgroup$
  • 2
    $\begingroup$ wouldn't this create a lot of duplicate information? $\endgroup$ – Franck Dernoncourt Jan 29 '17 at 21:27
  • 6
    $\begingroup$ @FranckDernoncourt, maybe. In case you haven't noticed, there is a tremendous amount of duplicate information on this website. For example, I'm pretty new to this site, and I can already see there are only a small handful of unique questions about subjects like multicollinearity and model selection in GLMs on this site, but that doesn't stop many users (including very high rep users or moderators) from supplying essentially the same answers over and over. I don't really mind this, but let's not pretend that non-duplication is sacrosanct on CrossValidated. $\endgroup$ – gammer Jan 29 '17 at 21:31
  • 1
    $\begingroup$ In the scenario you've described, this seems sufficiently rare (and, again, I'm new-ish so my perception may be off) that it seems like the most pragmatic solution is just to supply a combined answer to the combined duplicate question. And, just philosophically (e.g. using similar logic to that of academic research) I think of the combined question as distinct from the two individual ones. $\endgroup$ – gammer Jan 29 '17 at 21:44
  • 3
    $\begingroup$ (-1) No point to that, unless you're adding new info. - & even then I'd say there should be some benefit from the joint treatment of C & D. To be sure, there are many extant duplicates on the site, but we needn't let the best be the enemy of the good. $\endgroup$ – Scortchi - Reinstate Monica Jan 30 '17 at 11:34
  • 6
    $\begingroup$ One distinct question, one distinct thread remains an ideal. The more popular the topic, the more difficult it can be to track down the previous answer, even if you vaguely remember writing it yourself. Some people here have posted thousands of answers. $\endgroup$ – Nick Cox Jan 30 '17 at 11:45
  • 2
    $\begingroup$ @Scortchi, I agree with your comment. My point was just that, in many cases the joint treatment inherently adds something. This happens all the time in research. On this site, I could imagine an answer to a question about estimating intraclass tetrachoric correlations from a mixed effects probit model being mostly constructed from a cut-and-paste job from answers to questions about a) estimating intraclass correlations in a linear mixed fx model; and b) probit regression. And that seems ok. I didn't read the OP's links so I have no idea if that applies to his example. $\endgroup$ – gammer Jan 31 '17 at 4:05
  • $\begingroup$ (+1) I am not sure why this answer is getting downvotes. My reading of gammer's answer is that he is suggesting a "joint treatment" of C & D, like argued in Scortchi's answer too. I almost feel that @Scortchi's downvote was based on a misunderstanding. $\endgroup$ – amoeba says Reinstate Monica Jan 31 '17 at 21:57
  • $\begingroup$ @gammer: I feel your last comment qualifies your answer rather - you might want to edit it. $\endgroup$ – Scortchi - Reinstate Monica Jan 31 '17 at 22:49
  • $\begingroup$ @Scortchi, I'm not sure how I can edit it to reflect that, because that's what I meant by "distinct enough". Feel free to edit it or community wiki it if you want. $\endgroup$ – gammer Feb 1 '17 at 2:54
1
$\begingroup$

Gold-badge holders and moderators can now edit duplicate links (2017-03-03):

As of a few minutes ago, if you have a gold tag-badge for a tag associated with a question that's been closed as a duplicate, you can edit the duplicate links to replace, add, remove or re-arrange them:

edit link in duplicate banner

link editor

revision list

These changes are tracked in PostHistory to allow anyone to detect abuse, while the results update the PostLinks table so that existing logic continues to work (I mention this because some of you may wish to lean on SEDE for observing how this works out, once it gets updated).

$\endgroup$
0
$\begingroup$

What if a question is an exact duplicate of the conjunction of two other questions proposes another solution:

Moderator at Programmers meta recommended the way to handle conjunction cases in questions closed as duplicates as follows (bold font in quote is mine):

while the duplicate link is auto generated it is actually inserted into the question body so can be edited. It should always remain, but there's nothing to stop you (or anyone else) editing extra links into the list or indeed replacing the current one with a more appropriate one.

  • that's how a diff may look like for added extra links:

    http://i.stack.imgur.com/oHwlk.jpg

$\endgroup$
  • 3
    $\begingroup$ The problem w/ that suggestion is that it is very old. It was relevant at the time, but the way the duplication was marked in the SE system was changed sometime between Feb 2, 2013 (eg) & Feb 6, 2013 (eg). That suggestion would work under the old system, but not now. $\endgroup$ – gung - Reinstate Monica Feb 6 '17 at 21:58
  • $\begingroup$ @gung thanks I wasn't aware of these changes. $\endgroup$ – Franck Dernoncourt Feb 6 '17 at 22:01
  • $\begingroup$ -1 in light of what @gung wrote above. $\endgroup$ – amoeba says Reinstate Monica Feb 7 '17 at 13:29
  • 1
    $\begingroup$ @amoeba That's fair. I am leaving the answer so that someone else doesn't make the same suggestion. $\endgroup$ – Franck Dernoncourt Feb 7 '17 at 16:41

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .