Since this site frequently deals with larger statistical concepts, I frequently find that my explanations to users cover a lot of the same ground. For example, when I am answering a question on cross-validation, the beginning of my answer is always unique to the particular question, but at the end when I try to tie the particular example to the larger concept, I find myself saying similar things. I tell them the key to designing cross-validation is about thinking about how your new data you want to use the model on are different than your training data, and so you should try to split of your data for cross-validation so the cross-validation folds will mirror those novel aspects.

It seems like it might be more efficient to take the time to write out a few sentences illustrating this concept as eloquently as possible, and then copy and paste them at the end of the unique part of the answers. This would avoid rewriting it each time and have the quality of the explanation vary depending on other factors such as how much time I has to read over my answer. However, while the larger Stack Exchange community seems ok with copy and paste answers, Cross-Validated seems to frown on it.

So where would these partial copy and pastes fall on that scale? Are they ok because most of the answer is unique, or not because we only want unique content? I might consider linking to a previous answer with the repeated part, but usually the part I want to reference is just a few sentences at the end of an otherwise unrelated answer.

  • 4
    +1. A slightly newer MetaSO post at meta.stackexchange.com/a/211726/166924 looks helpful, too. The much more recent thread at meta.stackexchange.com/questions/210373 contains some remarks indicating SO is rethinking this five-year-old policy about allowing dup answers, having recognized it's leading to a lot of junk. – whuber Dec 19 '16 at 22:34
  • 12
    Cross-Validated frowns upon fully duplicate answers. "Partial duplicating" that you are describing here is arguably less bad and can be acceptable. If you copy-paste parts of your own earlier answer, I'd suggest to format them as a quote and link to the original answer. As a better solution I would suggest writing a perfect answer once and then linking to it in other answers whenever you feel necessary (without much or even any copy-pasting): e.g. by writing "See my answer here for more details" or some such. I often use such cross-links. – amoeba Dec 19 '16 at 23:07
  • 2
    @amoeba Thanks for your comment, and I agree that it is better to have once good answer once and link rather than copy pasting or rewriting. What I am more concerned about though are things that would never be an answer to a question on their own, but are often a good guide on how to derive an answer. These are usually a last few sentences to paragraph at the end of another 2-3 paragraphs of answer. It seems less appropriate to link to an answer that is 3/4 unrelated to the asker's question just for a few sentences at the end that are. – Barker Dec 20 '16 at 0:22
  • 7
    +1 I sometimes paste short parts of other answers when relevant (e.g. examples, definitions) with providing link to the full answer. In my opinion this is better then needing OP to jump over multiple links to get the answer to his question. If exactly the same answer applies to different questions, then I vote to close them as duplicates rather then pasting or re-writing the whole answer. – Tim Dec 20 '16 at 8:50
  • 1
  • Hi Barker, I see your point. I think it could be useful for this discussion if you could provide an example of such "last few sentences" that you repeatedly feel inclined to include into your answers. This could make the discussion more specific. Can you post it here as comment? – amoeba Dec 26 '16 at 14:15
up vote 3 down vote accepted

If it is really two sentences within a longer answer, it should be OK. No need to paraphrase yourself every time for what is actually a citation.

But having known similar situation on other sites, I would suggest a different approach instead: self-answered canonical questions.

These are really useful when there is a confusing concept which novices often get wrong. For example, let's say that somebody comes with a very concrete problem about a t-test, and you help him straighten his calculation, but also notice that he has no good reason to assume equal variances. So at the end of the answer, you also advise to choose a different test instead, and remember that this is a situation which crops up every now and then.

You can of course reuse some nice explanation you have written as a part of another concrete problem. But you can also write a question, "How do I know if I can use a t-test" and write up an answer which gives a good background about why you shouldn't use a t-test when the variances are unequal, when is it OK to assume equal variances, and what to do if you cannot assume them. Then link from your concrete answer to this well-packaged bit of basic knowledge whenever you need.

This strategy has its drawbacks of course, and is not the one to be used in all cases. Sometimes copy-pasting the couple of sentences is the better thing to do. But keep in mind this alternative too, it can be very fruitful.

You must log in to answer this question.

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