11
$\begingroup$

I know that people here are very squeamish and punctilious about editing questions of others, and rightly so (see editing is not rewriting). But here is one particular situation that I would like to discuss.

Here is a well-upvoted question with an accepted well-upvoted answer: How does centering the data get rid of the intercept? The question is, as the title indicates, specifically about linear regression. However, @ttnphns's answer also explains what centering does for PCA (and provides a very helpful figure). As a result, new questions about centering in PCA get closed as duplicates of that one.

The problem is that there is no way to guess from the title of this question that there is anything about PCA inside. So my intention is to edit the title to e.g. "How does centering the data get rid of the intercept? How does it affect PCA?"

But the OP was not asking about PCA. Is it okay under these circumstances to ignore OP's intention and proceed with broadening the title?

| |
$\endgroup$
  • $\begingroup$ One week after posting this meta question, with nobody saying that this would be a bad thing to do or downvoting my post (currently 7 upvotes, 0 downvotes), I went ahead and made the suggested title edit. $\endgroup$ – amoeba Dec 22 '14 at 16:07
2
$\begingroup$

A search of type [edits][title] on Meta Stack Exchange and Stack Overflow Meta did not help much.

My two cents. I'd say we should be cautions to edit a title to encompass (all) existing answers. The recent overfitting question asked for real world examples and it collected answers out of this scope. Not sure if the title should be edited in that occasion.

On the other hand, editing the question's example seems perfectly fine.

  • the OP mentioned PCA in the question body.
  • the answerer of that question tagged it with .

You would be adapting/improving the title to what already exists in the question as a matter of content.

| |
$\endgroup$
  • 2
    $\begingroup$ Re overfitting question: I think that "real world" can be understood more broadly than "actually taken place" (as I argued on the related meta thread) so I think that title is fine. Come to think of it, I would put it in plural and remove quotes: What are real-world examples of overfitting? $\endgroup$ – amoeba Dec 15 '14 at 16:57
  • 1
    $\begingroup$ Re my example about PCA: well, the question body does mention PCA but in a different context (I believe). But you are right, editing in the [pca] tag was already a potentially (??) controversial edit along the lines of what I am asking here about. $\endgroup$ – amoeba Dec 15 '14 at 16:59
  • $\begingroup$ By the way, this is not an isolated case, I could come up with other examples. $\endgroup$ – amoeba Dec 15 '14 at 17:00
  • $\begingroup$ Re overfitting question: for example, I would not change the title there to What are examples of overfitting? (removing the real-world part). But keeping on-topic here, it is not ok to change the title to encompass content embedded in one answer, if the answer is off topic to the question. Clearly, that was not the case in the example you cited. $\endgroup$ – Andre Silva Dec 15 '14 at 17:05
  • 1
    $\begingroup$ I might include some other examples later, but for now assume that the OP had not not even mentioned PCA in the question's body. Then it seems to me that your last sentence would not really apply. What then? Just to be clear: thanks for your answer, I appreciate it, but of course I am still hoping to collect opinions of other users as well. This is a discussion thread, I don't think there is one "correct" answer. $\endgroup$ – amoeba Dec 15 '14 at 17:16
  • $\begingroup$ Sure amoeba, that is fine. I suppose if the OP had not mentioned PCA, then the proposed edit would not make sense. In such situation, probably ttnphns's answer would not be there either. $\endgroup$ – Andre Silva Dec 15 '14 at 17:31
  • $\begingroup$ I disagree completely. The OP reads: I keep reading about instances where we center the data (e.g., with regularization or PCA) in order to remove the intercept. As I understand it, this refers to centering in regression, and PCA is mentioned as one possible way to center the variables. The question though is about what happens with regression after centering, not about what happens with PCA after centering. So this mentioning of PCA is a red herring. Still, ttnphns's answer might have been nevertheless motivated by it of course. $\endgroup$ – amoeba Dec 15 '14 at 17:35
  • 2
    $\begingroup$ @amoeba When I edit I try to take particular care to respect the intent of the OP. But on a years-old thread with a long-departed OP, now the thread is mostly of use search engine arrivals, maybe it's a silly hang-up of mine. For Qs like "How do I do X in SPSS/R?" the on-topic meat of the Q is usually "How do I do X?" so I don't mind changing such titles. Other times it's unclear whether X should be construed in a wider way that includes Y too. Particularly as titles don't always adequately summarise question bodies! Ensuring title reflects body seems wise to me. ... $\endgroup$ – Silverfish Dec 15 '14 at 21:00
  • 1
    $\begingroup$ ..so "You would be adapting/improving the title to what already exists in the question as a matter of content" summarises my opinion (+1). Re overfitting, I agree with @amoeba. A restrictive reading of the title suggests a narrow scope, but it can be read as "why/where in the real world, might an analyst overfit a model?" The body says "I need help as to how to come up with a real-world example that applies to overfitting", where "come up with" seems more likely to mean "create my own example" than "find a published paper". Only OP can edit to correct vagueness as only they know their intent $\endgroup$ – Silverfish Dec 15 '14 at 21:10
  • 2
    $\begingroup$ NB answers are often wider than Qs on CV, Alecos Papadopoulos blogged this, generally favouring "wider" answers. Also, non-expert often ask questions which receive answers addressing their deeper intent. "How do I test X?" often gets "Your data are inappropriate for X because of reason R. But if you want to do X because your underlying goal was A, have you considered test Y?" Yet the title concerns X. Is "X vs Y" implicitly existing in the Q? Should it be edited in? Hmm. $\endgroup$ – Silverfish Dec 15 '14 at 21:18

You must log in to answer this question.

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