R-specific StackExchange site or greater integration of R community within CV?

Question

There has been discussion over how more of the R community could be brought into the StackExchange fold. (See also this Question on SO Meta.)

If I were to sum-up feelings expressed thus far, we are split between proposing a new, all encompassing SE site for R (R.SE) or finding a bigger home here on CV. @mbq suggested that much of what we might do on R.SE would be perfectly at home on CV.

Ideas for the R.SE site include (but are not limited to):

• Using/deploying R in a variety of situations (teaching, research etc)
• Using the various interfaces to R, questions on RStudio, RevolutionR, Rapache, ... etc...
• Domain specific Qs on R packages
• Using R in a very general sense (not programming per se)
• (anything covered by R-Help, R-Devel, and the SIG list, but we might off-load statistical Qs to CV and R programming to SO, where they were clearly a better fit.)
• Allow the more open style of Q

What do the users/mods of CV think about this? Would such questions find a welcoming home on CV or would we run the risk of drowning out the statistical Qs?

Any comments gratefully received, either way. At the moment we are deciding upon the best course of action, so anything you do say will help inform that.

Answer 1

First, some context, as I see it: I think of CV being like the Physics, Electrical Engineering, and other knowledge-domain StackExchange sites. I think of SO being a big programmer / developer / software-engineer-oriented site, i.e. people close to the coding and implementation of stuff.

R is a programming language that is accessed from multiple environments -- the standard R console, RStudio, Revolution R, Deducer, JGR, ESS, etc. -- and multiple R<-->Other language/software connections (e.g. rJava, JRI, Rcpp, RExcel, RCOM, etc.). As I stated on the Meta page, I think that people may solve problems in R, but usually they're using R to solve a problem from some other context (e.g. regression, visualization, etc.). Because of that and other differences from working in C, Java, etc., it's not a typical cultural fit for SO - it's better to have a separate site like the Ubuntu, Linux, and other environment folks do.

To me, grafting R onto CV is like grafting Matlab onto the Electrical Engineering site, except that R is a far more pervasive beastie and is evolving at a far more rapid rate than any other language. Personally, I would like to use both sites (i.e. CV and a new R.SE site), and encourage questions to be submitted to the right site.

Answer 2

Sometimes it helps to get a different perspective. What if, for instance, we were to replace every occurrence of "R" in this discussion by "SAS"? In principle that shouldn't change the validity of any of the arguments pro or con. To see what this would entail, think about what our site would be like if it were inundated with questions like this one.

Vertically joining data in two (or more) data sets using proc sql. If each data set represents data gathered in different years data_2000, data_2001, data_2002, ...

If one wants to take a column from each file, but add a column populated with the year each row was taken from, creating the column is easy enough, but I don't know how to populate the values for each row.

I'd like to use a macro, give a list of files and years, let the macro generate the proc sql code to join given columns, et cetera

Frankly, I'm not enthusiastic about that prospect. I don't care whether such questions come from R, SAS, or whatever my favorite statistical computing platform happens to be at the moment: they are pedestrian and plainly uninteresting. They belong somewhere else so that we can focus on statistics and data mining here without such distractions.

Answer 3

This is my first post on CrossValidated or Meta-CrossValidated, so I don't feel I'm a stakeholder here. FWIW, I'm trying to learn R and statistics - but I come from an implementer's perspective.

The situation right now is that R questions are split across StackOverflow and CrossValidated, with the unfortunate result that good questions go unanswered and good answers go unseen. See, for example, this StackOverflow question from today, which duplicates this CrossValidated question from four days ago.

I don't know what you/we want to do with these "pedestrian" questions in the long term, but in the short term, I'd suggest that it might be better for everyone if folks here were "harsher" about kicking plain R questions, like the one above, over to StackOverflow.

I don't think the argument that "many statisticians using R who would be unlikely to want to keep up with a different site" is valid here, simply because if they aren't looking on StackOverflow, they're already missing (very roughly) two-thirds of the R traffic.

There are probably some nominally R-centric questions that are really statistics questions couched in R terminology - you might want to keep them.

An alternative, outside-the-box solution would be if StackExchange provided a way for certain tags on one site to show up in searches for that tag on certain other sites. This may be fraught with downsides and implementation difficulties, though.

Answer 4

Moving R away from stats would be a significant mistake in my view. There are many statisticians using R who would be unlikely to want to keep up with a different site.

Answer 5

I still think that inviting R-help here is a better idea than it seems; both mailing lists and SE sites are really made to serve communities, not topics. This site is for general data science community, R-help is for data science community using R.

As can be seen from R-help archives, it is discussing very similar topics as we do here; yet this community is 1) large and very active, 2) certainly more engaged in R programming issues than ours.

1) will result in some overwhelming (my approximation is that it will be an equivalent of merging CV with a second site of more-less the same size). However, those people are statisticians and data miners, and there is a great chance they will start to ask more core questions and it finally only end in a bigger site with more-less original topic distribution.

2) might lead to a serious topic drift; however there is a solution, namely maintaining the existing SO/CV split of more "sciency" and more "programmy" R questions.

The smaller in size but also important issue are some discursive and big-list questions that are in fact against general SE rules. SO is very broad and thus SO mods will always kill such threads; here we can at least give it more thought.