3

I noticed this morning that we now have a tag, with three questions and a large number of pending suggested edits to add questions to this tag.

This tag doesn't talk about an area of cognitive sciences, psychology, psychiatry nor neuroscience and to me, it feels like a meta tag.

Do we like the "correlation" tag or do we burninate it?

| |
  • Not looking good for the tag so far @Skippy... +6 to burninate, -1 to keep it... – Josh Aug 19 '13 at 21:31
  • It is now +7,-1 to burninate, on reflection and reading all the arguments here, I can see that this is an unnecessary tag. Maybe down the track, but as it stands, not necessary at the moment at least. – user3180 Aug 24 '13 at 6:14
7

I suggest we burninate it.

The editor in question seems to be searching for the term "correlation" and applying the tag to all those relevant questions, which is kind of an indication the tag isn't needed at all, even if it were to be a proper tag.

It's more of a meta tag the way they are currently applied. It could be a valid tag if the subject is more related to e.g. statistics in general.

| |
  • 1
    @Skippy I was referring to helping out with statistics in general, rather than applied to one specific instance. E.g. "What are common ways to find correlations between self-reported Likert scale values?" When posting this I asked myself the question: "Would it be useful to look for all questions related to correlation?" I don't see who would be interested in that, as opposed to e.g. intelligence as in this question. Read through the meta tag and let us know where you disagree. – Steven Jeuris Aug 15 '13 at 8:38
  • @Skippy chat.stackexchange.com/transcript/message/11116688#11116688 But oh well, ... what's done is done. :) – Steven Jeuris Sep 7 '13 at 12:08
4

From what I can tell, the tag already existed in a couple of questions, and @skippy was thinking of ways to help the site by adding the correlation tag wherever it seemed relevant. I appreciate the intention and in particular @skippy's been doing some great things on the site in terms of questions and answers.

However, in this case, we have a tag that I think should never have been created in the first place. Correlation concerns the relationship between two variables. In the limited sense of Pearson's correlation, we are talking about the linear relationship, but in the broader sense it could be almost any bivariate association. Science is all about relationships between variables, and there's not much shared by two questions because they are merely concerned with the relationship between two variables, most of which never mention the word "correlation" explicitly. Many questions on this site are concerned with the relationship between two variables. In fact, there's even a question of what is meant by correlation (is it just for observational studies; what about studies that look at the relationship between variables and how a third affects that relationship). Thus, I agree with Steven. Let's get rid of the "correlation" tag.

If there was any argument for having a correlation tag, it would be for questions that specifically pertain to the "correlation" as a topic. For example: What does a correlation mean? Why use Spearman's versus Pearson's correlation? When can a correlation be interpreted as causal? That said, this site doesn't get many questions like that so I think the statistics tag is probably adequate.

More generally, as our site grows and we have more moderate rep users who have permission to create new tags, we are going to get an explosion of infrequently used tags. Thus, the bigger question is about how much we want to curate tags. I guess the main lesson from this experience is if a users wants to edit tags on mass they should probably post a meta-question first to check that the community is okay about this. If a tag needs to be post-hoc applied to 3+ questions, there may often be a reason for why it was never applied in the first place.

| |
  • 3
    Great. Feel free to contribute to wiki tag edits and excerpts. It's great when new users start getting involved with that next level of the site (i.e., tags, wikis, edits, revisions, flags, etc.) – Jeromy Anglim Aug 15 '13 at 8:22
  • 2
    @Skippy, as Jeromy states in this answer, if you are going to do mass edits, it's worthwhile discussing why on meta first. I think that's a pretty good guideline. As to any other edits, this post is a good place to start reading on how to do them. Again, feel free to join in on that discussion. We only have one answer there, so different views are more than welcome! – Steven Jeuris Aug 15 '13 at 9:29
3

Just as a side note: I think that the tag wiki entry for might be wrong in its current form anyway:

The link between two or more things. The mathematical relationship between entities. So the change in one variable resulting in a change in a separate variable.

It seems to me that the last sentence is a description of causation, not correlation.

| |
  • 3
    That's an important point. While the tag is there, the description should be changed. Correlation describes the relationships between two, not more variables. And it sounds a bit strange, to speak about entities in the context of correlation, although it may not be wrong. And I also think that the last sentence can be interpreted in terms of causation. As @JeromyAnglim said, there are different types of correlation. Especially because correlations are so commonly used in scientific research, I feel it takes away from the credibility of this site if wqe don't have a sound description. – Jens Kouros Aug 19 '13 at 7:35
2

Questions that need a correlation tag are off-topic here and should go to stats.SE.

| |

You must log in to answer this question.

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