Currently we have the following regression tags: regression, linear-regression, logistic-regression, nonlinear-regression, stepwise-regression, ridge-regression, dynamic-regression, meta-regression. This is perfectly normal, since there are a lot of questions related to the regression and separate tags makes a lot of sense.

I would like to propose however to remove multiple-regression tag, since usually all the regressions are multiple.

I also propose standard formulation for all the specific regression tag wikis:

[Insert regression type] regression is a type of regression, where [insert some characterisation of the particular regression type]

The example for linear regression would be:

Linear regression is a type of regression, where regression function is linear. This is the most widely used regression type.

I've already edited linear, logistic and dynamic regression tag wiki's, the edits will appear when they are approved.

Update I see that linear regression tag was merged with regression tag. This seems a good thing. It would be also good to merge the multiple-regression tag with regression (by merging I mean making synonym or something to that effect). I looked at the questions tagged with this tag, they all mention linear regression.

  • $\begingroup$ @mpiktas Ah, it appears I already validated two of them before seeing this thread. Let's wait and see what other users think of your proposition. (We can modify the tag excerpt afterwards.) $\endgroup$ – chl Apr 22 '11 at 7:47
  • $\begingroup$ @chl, that was my intention in part, because there is no way for others to see how the proposed wiki looks when modified. Currently only 10 users can approve wiki, i.e. see the proposals. $\endgroup$ – mpiktas Apr 22 '11 at 8:07
  • $\begingroup$ What is to distinguish the regression tag with the more specific tags? $\endgroup$ – Andy W Apr 25 '11 at 13:05
  • $\begingroup$ @Andy W, good question. $\endgroup$ – mpiktas Apr 26 '11 at 8:30
  • $\begingroup$ @Andy W, can you clarify whether you mean that other regression tags are unnecessary, or something else? $\endgroup$ – mpiktas May 3 '11 at 7:40
  • $\begingroup$ @mpiktas, I'll try to add clarification sometime today in an answer. I think they are necessary and useful, but I'm concerned about the scope of the "regression" tag versus the more specific tags. $\endgroup$ – Andy W May 3 '11 at 12:38

I believe the multiple-regression tag has a purpose, because there is a clear distinction maintained in textbooks and people's minds between "simple" regression models and "multiple" regression models. A regression problem becomes overtly "multiple" when one has to grapple with the mutual relationships among the independent variables. To clarify this, I have created a wiki entry for multiple-regression to this effect. As always, your suggestions about improving that are welcome.

It is likely that the distinction I am highlighting has not consistently been maintained by previous users of the multiple-regression tag. The fix is to retag such questions whenever you run across them.

Alternatively, one could argue that the combination of the multivariate and regression tags achieves what my multiple-regression proposal does. I think such an argument is valid and am interested to hear what people find more useful or natural.

  • $\begingroup$ Do you think it is an appropriate burden on the community to have to continually retag the questions? Is it problematic that for many of the users of the site the tags will not have any immediate recognizable distinction? $\endgroup$ – Andy W May 4 '11 at 12:16
  • $\begingroup$ @Andy Those are valid concerns. My impression is that this is the status quo: many questions are inappropriately tagged at the outset. That they need retagging becomes evident in light of elucidations in comments or subsequent replies. Rather than retagging being a "burden" on the community it becomes an ongoing tool to clarify and codify the questions and answers. As far as "immediately recognizable distinctions" go, it's the purpose of the wiki entry to make those distinctions. See the multiple-regression entry, e.g.. $\endgroup$ – whuber Mod May 4 '11 at 13:28

Initially I had thought that maintaining a separate tag for regression distinct from the other tags (e.g. multiple-regression, linear-regression, etc.) would be worthwhile. Mainly because I feel there are questions that are related to regression, but aren't intrinsically related to the more specific tags. For instance this question (whether a predictor with greater variance is better) is not inherently related to linear-regression, while this question (Changing null hypothesis) is. But that being said, I suspect to maintain any type of fidelity between the regression and the linear-regression tags it would take heavy moderator involvement forever.

I would like to see as much discrimination between the tags as possible though. I see 10 questions have both the regression and the logistic-regression tag, and I think all of them could effectively drop one of the tags. None of the other regression tags are used frequently enough to be concerned, but should we attempt to only tag questions with one of the regression tags in most situations?


You must log in to answer this question.

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