# Independence tag

I think the tag should be split in two (or something): One for independent data the other for independent variables. These are often confusing.

• @AndreSilva, I think it does make sense to edit the tags for questions that are about IVs, but use [independence] instead of [predictor]. I would only say that it's best not to do all of them at once. – gung - Reinstate Monica Jul 2 '13 at 14:05
• @AndreSilva, what I mean is: if they use [independence] instead of [predictor]. – gung - Reinstate Monica Jul 2 '13 at 14:49
• @Peter, Do you think the tag dependence (10 uses) is a synonym of non-independent (55 uses)? I was trying to check this out and I remembered this thread of yours (maybe you have come across with this too). Thanks. – Andre Silva Aug 4 '13 at 14:46
• They do seem alike. "non-independence" seems awkward. – Peter Flom Aug 4 '13 at 21:08

In regression, variables are rarely independent. If you have a balanced experiment design, they can be made independent and uncorrelated (assuming you have no missing data). In pretty much every other situation, including every observational study (and that's what people usually come to this site with), the regressors are correlated, and hence cannot be said to be independent. While I am totally fine referring to the response variable as the "dependent variable", I deliberately avoid the parallel, and always refer to the regressors as "explanatory variables".

If we can help the utterly confused readers of the site, we should do this. I am in no position to enforce my idiosyncratic terminology use on everybody, but I would suggest rethinking what the independent variables may refer to.

Oh yes, the factors that come out of principal component analysis are uncorrelated (and hence independent if they are normal). And there's the method of independent component analysis, properly tagged ica.

This is a reasonable point. I just edited the tag's wiki & wiki excerpt to indicate that it is not for independent variables.

• Should I create a independent variables tag? – Peter Flom Jun 29 '13 at 19:25
• @PeterFlom, I would say no, unless you just wanted to make it a synonym. We have a predictor tag, which is intended to include IV. – gung - Reinstate Monica Jun 29 '13 at 19:44
• @gung Good addition to that tag – Glen_b Jun 29 '13 at 23:41
• @Peter Good idea. I created independent-variables as a synonym of predictor. – whuber Jul 10 '13 at 14:25

This is a bit of a repeat of the answer I posted here:
Error tag is ambiguous

...but we should really consider ending the use of the ambiguous tag and moving to new ones.

The new tag terms should not fight users, though. Many, many people use the term "independent variable", so you will have a constant battle on your hands if you try to enforce the exclusive use of predictor. "independent variable" and "predictor (variable)" are synonyms, so the tags should match that.

I propose:

• [independent variable] (or [independent variables]), and
• [independent data]

Doing it this way will make the re-tagging job easier, leaving [independence] as the tag to mean "I need a new tag".

Most importantly, whenever users start typing in "independen...", they will find a tag that matches what they want.

• That makes sense to me – Peter Flom Jul 5 '13 at 10:58