What is meta? ×
Meta Stack Exchange is where users like you discuss bugs, features, and support issues that affect the software powering all 127 Stack Exchange communities.

A sizeable R community has grown up around the [r] tag on SO. However, we are increasingly seeing a large number of non-programming related Qs being posted to the site. Also, some of the more open-ended or no-single-answer Qs that are highly thought of by the [r] users are being closed by SO Mods as they transgress the "What not to ask..." parts of the FAQ.

Whilst R is a programming language, it is far more than this, with people using high-level R functions and packages to do data analysis for both work and play. There is an active and vibrant R community that extended well beyond the concept of programming.

Some of us in the [r] SO community have been discussing the possibility of getting together a proposal for Area 51 for an R-specific Stack Exchange site. This site would embrace the entire R community and the myriad uses to which R is put. Statistical Qs would still be best housed on CrossValidated and R programming Qs could remain on SO, but everything else could be off-loaded onto the new site, as well as welcoming a wider selection of Qs. In some senses, it would be a little like the relationship between SO and Programmers; they coexist happily with different remits.

It has been noted that there are few, if any, language-specific sites proposed on Area 51. Is this because they would be inappropriate and have been deleted? Would such SE sites be welcomed if we could garner sufficient support from the R community at large? The proposed site would be very much in the spirit of environment-based sites such as Drupal, Wordpress, Ubuntu etc.

share|improve this question
1  
you may want to edit the title to explain that this would be a programming language site, not a spoken/written one. –  tombull89 Aug 1 '11 at 17:41
1  
@tombull89 Avoided "Programming" at first in the title because R is more than a programming language per se, but point taken, thanks. –  Gavin Simpson Aug 1 '11 at 17:44
2  
I've not learned the Meta answer "culture", but I'd like to offer an extended comment (despite the poor formatting). I'd recommend expanding the concept beyond just the language and think in terms of a whole environment. 1. Language questions often ignore the compiler/intrepreter; however, people often fiddle with the R interpreter (e.g. by compiling or linking to other sources). 2. There is 1 language, yet multiple R interfaces exist: Revolution R, RStudio, ESS, Deducer, etc. 3. Users are actively contributing to the development of the infrastructure itself. –  Iterator Aug 1 '11 at 18:40
1  
To continue the comment: 4. Culture: there is a cultural difference between using R to solve a problem and solving a problem in R. Users come to R because it helps them solve another problem. That's not true of typical programming languages and represents a cultural problem for StackOverflow moderators. 5. Moderation: I believe that SO has a weakness relative to forums in general in that I do not see that topic expertise is related to moderation activities. A separate site allows for better matching of site users and site moderation practices. –  Iterator Aug 1 '11 at 18:42
1  
I think @Iterator makes a lot of valid points. You should probably go ahead and put all of this into an R-Environment proposal on Area 51. –  Bill the Lizard Aug 1 '11 at 19:32
    
@Iterator Yes, indeed. The scope would be everything and anything R, writ large! I'll look to garner suggestions for the proposal including the points you raise. The Q here was mainly to sound out the language-specific issue, but I am glad it has teased out useful info from yourself and Bill. –  Gavin Simpson Aug 1 '11 at 19:37
    
If something will be drawn up, then my second point is missing the standard R interpreter. :) (To say nothing of R CMD, RApache, littler, Rserve, JGR, and all the R-<your language here> interfaces.) –  Iterator Aug 1 '11 at 20:03
2  
I think the title should be changed to reflect that this is about R specifically. I think few things that are discussed here can easily be applied to other languages. –  Joachim Sauer Aug 2 '11 at 15:25
    
In the future, perhaps we can avoid describing R as just a language? –  Iterator Aug 2 '11 at 18:37
add comment

5 Answers

up vote 19 down vote accepted

Statistical Qs would still be best housed on CrossValidated and R programming Qs could remain on SO, but everything else could be off-loaded onto the new site, as well as welcoming a wider selection of Qs.

That would leave a tiny number of potential questions.

It has been noted that there are few, if any, language-specific sites proposed on Area 51. Is this because they would be inappropriate and have been deleted?

Yes. The thinking is that a site dedicated to only one specific programming language would probably be too small to attract a critical mass of experts to answer questions. It would miss out on the benefits of the larger SO community. Most programmers know more than one language, so are able to answer questions in many different SO tags. If you narrow a site down to just one language, that benefit goes away.

share|improve this answer
    
I disagree about the "tiny" number of potential Qs. Head over to R-Help and other SIG mailing lists and see how many Qs are posted there that are not directly about statistics. There are 3000+ packages on CRAN, how do I use package X to do Foo would be appropriate Qs but not really on topic for SO or CV, though the mods at CV suggest the scope of what is on topic there could be relaxed. It has been genuinely suggested that SO/CV or a SE site could replace R-Help, for example. –  Gavin Simpson Aug 1 '11 at 17:57
4  
A counter to the SO site, is that we are not being allowed to have the useful, informative, yet more open-ended Qs around R because Mods are closing them as OT. Several of us feel this is a bit heavy-handed, but the FAQ is the FAQ. –  Gavin Simpson Aug 1 '11 at 18:00
    
@Gavin: "how do I use package X to do Foo" sounds like it would be on topic for SO. Regardless of my opinion, you can still open a proposal on Area 51, but I'd suggest you come up with a lot of good example questions that would be off-topic for SO and CV. –  Bill the Lizard Aug 1 '11 at 18:00
4  
@Gavin: You'll need to link to some of those questions. I've only seen one recently and it was closed for being a poll, which would have happened on any SE site. –  Bill the Lizard Aug 1 '11 at 18:01
    
thanks for your comments. Will see what else comes up here and then see what the people in the R chat room think about the next step. Cheers! –  Gavin Simpson Aug 1 '11 at 18:02
    
Here's one that was closed as a Poll. stackoverflow.com/q/6796490/429846 One commentator suggested asking on Programmers (but noted the poll issue too) This is a great Q with lots of useful info in a few Answers. (I hesitate to mention) stackoverflow.com/q/1295955/429846 (for fear of getting that closed too), and stackoverflow.com/q/6246762/429846 the closing of which annoyed quite a few of us (see comments) –  Gavin Simpson Aug 1 '11 at 18:07
3  
And another one that we really had to fight for to keep on SO: stackoverflow.com/q/5963269/602276 –  Andrie Aug 1 '11 at 18:59
2  
Is it possible to have tag-specific moderation within SO? If so then it is pointless to have non-R users moderating topics that R users think are on-topic and worth answering. This is a cultural & expertise matter, but the question, in my mind, is whether or not a different site needs to be developed or if new functionality is needed on SO to ensure that moderation is done by people who know both the topical matters and the SO gestalt. Users are fluid: a site will support their interests. Can that site be SO or a separate site? I think it can be SO, else migration is necessary. –  Iterator Aug 1 '11 at 19:24
5  
@Iterator: No, it's really not feasible to have tag-specific moderation within SO. –  Bill the Lizard Aug 1 '11 at 19:30
2  
@Bill Thanks. Absent that kind of functionality, then the best way to have such moderation currently seems to necessitate a separate site. Kind of a heavy hammer, but StackExchange does make it easy to do. –  Iterator Aug 1 '11 at 20:01
3  
@Bill - can you clarify your point earlier regarding the poll type questions and being closed on any SE site? From reading the comments and trying to read between the lines, I think others (Gavin, Iterator, et. al.) are suggesting / hoping that the types of questions linked to above would be allowed on the theoretical R-SE site. If the newly minted moderators of said site are tolerable of such questions, would one of the head honchos from the main SE site step in and ensure order? –  Chase Aug 2 '11 at 4:36
3  
@Chase: I don't know of any SE site that accepts poll-style questions. The software really isn't ideal for it. I have argued in the past that if a poll-style question does not break any of the other rules of the site (regarding quality and subjectivity) and does not exist elsewhere on the Web, then it should be allowed to stay. That met with a tepid response from the community, so I can't guarantee that the SE management would go along with allowing them. –  Bill the Lizard Aug 2 '11 at 11:19
    
@Bill : your point of view here is contradictory with your point of view at meta.stackexchange.com/questions/100617/… : either these questions are allowed on SO, or there is room enough for an R.stackexchange site. The argument there won me for the R.stackexchange site. I've seen a discussion like this come up about every week on questions in the R tag. –  Joris Meys Aug 2 '11 at 12:10
8  
@Gavin As one of the three CV mods, I may be in the minority but I must disagree that "the mods at CV suggest the scope of what is on topic there could be relaxed." I would argue to the contrary: plenty of R-focused questions have been posted that I would prefer to have closed, yet did not because they have no suitable home elsewhere. An R-focused SE site would solve that problem, although it would raise a new problem of where to direct questions. Briefly: if a Q is about how to do statistics or visualization, it belongs on CV, even if the preferred implementation environment is R. –  whuber Aug 2 '11 at 14:59
1  
@whuber sorry, that was a slip of the keyboard. I should have said one CV mod. I have been careful (in other regards) to not assume that this was the general will of the CV mods or community. As one of the latter, I personally do not see CV as being the right home for the R community. –  Gavin Simpson Aug 2 '11 at 15:02
show 3 more comments

I've deleted my previous answer and add another, as I recently changed my mind.

As Joachim noted, the R community needs a place to be a community. Or more specifically, need a Q&A site about R where the rules can be determined by the R community itself. As recent discussions have shown, the specific nature of R as both a programming language and a statistical tool doesn't fit the current SO policy well.

Some of the issues :

  • development in R is quite different from developing in a programming language that does not function as a statistical analyzer/visualizer as well. Yet, questions about tools are not according to SO standards.
  • R has different standards for different tasks. It is interesting to know which formats/ standards are useful in which cases, but again, this will be closed as subjective and argumentative.
  • As R is also about statistics and visualization, the rigorous policy of SO results in quite a few R questions been deemed off-topic.
  • R-users are far from all programmers. In fact, the majority are not programmers at all.

So hence, it seems logic that, exactly like what happened for latex with http://www.tex.stackexchange.com, the R community finds a new home.

share|improve this answer
    
From what I read, I'm not exactly sure if a separate SE site will solve all of the problems, but it's probably a good idea. The TeX comparison sold me. (I wish I could downvote my answer so this goes further up ;-)) –  Joachim Sauer Aug 2 '11 at 15:18
add comment

I want to offer an additional answer with a thought experiment that might ache the heads of most folks focusing on the minor issues of what's on topic and what's not, especially relative to typical questions on SO.

Suppose I ask the following: "I am developing a clinical protocol package in R that will need to be reviewed by the FDA. What references and guides exist in order to ensure that this code will pass muster?"

There are correct answers. There is the FDA website, but an R-knowledgeable user will also recognize that saying one is using R is not like saying that one is using C with gcc 4.3 or Java version 6.666. It goes to questions like the packages used, the selection of random number seeds, documentation of all data transformations, etc. One would also have to address 21 CFR Part 11 certification, and so on.

Honestly, there is a large community of R and SAS users who know this stuff. They're often called biostatisticians. These are not your typical "enthusiast programmers"-style of questions. And this stuff affects the lives of people you know. No kidding. Botched data processing or software can lead to problems once the pill production starts.

Is there a place on SO for this question? Not really. Could an R.SE site handle it? Without breaking a sweat.

Note: industries like transportation, infrastructure, finance, etc., are also important, but I wanted to go with the most scrutinized & most insufficiently scrutinized domain first, these happen to be one and the same.

share|improve this answer
    
Can CV handle this? Yes. –  mbq Aug 10 '11 at 9:15
    
With all due respect, as I realize you're a moderator for CV, I disagree. I disagree just because of my observations over the last month or so of using SO and CV. CV should be able to handle this, and perhaps will be able to in time, but at the moment it looks to me like CV has more of an amateur than professional userbase. –  Iterator Aug 10 '11 at 11:48
    
(continued) The emphasis on CV is on statistics, not software, and vice versa for SO. However, the level of programming acumen on SO for R seems to be several notches higher than the corresponding statistical acumen on CV. The issue here with FDA clearance is more related to programming acumen, in any case. –  Iterator Aug 10 '11 at 11:49
    
(continued) I would sincerely like to elevate the statistical acumen of both answerers and askers on CV so that it could handle the statistical aspects of questions on FDA review of, say, clinical trials. However, that's different from grappling with both software engineering and statistical certification. Elevating CV on software engineering might be like pushing a string, because of the topical mismatches. –  Iterator Aug 10 '11 at 11:53
1  
CV is for people working with data, so this Q is in scope. CV is very young and in still in development, so it is not a good idea to judge it from history -- maybe you should try asking about FDA review? –  mbq Aug 10 '11 at 12:19
1  
Thanks for replying. Well, at the moment I wouldn't need the question answered. :) In any case, I will continue to contribute to CV. However, please re-read the question carefully. FDA review of software and devices isn't the same as clinical trial data review. I believe that for data review CV is definitely the most appropriate SE site. –  Iterator Aug 10 '11 at 12:51
add comment

Disclaimer: I have very little experience with R. I've dabbled and occasionally check questions out of curiosity.

It looks to me as if the R community needs ... well, some place to "be a community".

SO (and SE) are great sites for Q&A and that part of a community seems to be well-served by SO, even for R. (Edit: Joris' new answer convinced me that this may not be entirely true).

What SO can not provide is to be the place where subjective, discussion-based interaction happens. And I don't believe that a R-specific SE site would solve this, as the it's not just the culture of SO that tries to prevent this, but also the technology behind SE, which is not targeted towards discussion.

Believe it or not, but I think an old-school forum and/or mailing lists might be the correct approach here: Keep the technical "how do I do X in R"- and "why does this R code not do what I want?"-question on SO, but move "How should I approach the design for X"- and "I want to propose a new packaging format"-discussion to that place.

share|improve this answer
    
I've not yet come to understand how the SO technology is inadequate. It's a database, some GUIs, and some very primitive analytics. While it may not be Turing-complete, it doesn't seem how it can't be used for multiple purposes. –  Iterator Aug 2 '11 at 14:22
    
@Iterator: the question/answer architecture is really bad for discussions: you can't easily follow a train-of-thought (there is no threading), you can't easily "answer" one answer with another one (there's no relation between answers), ... All in all the entire system is designed for providing correct answers to objectively answerable questions. Human interactions that involve discussion, trains-of-thoughts between multiple participants and similar useful stuff is not easily represented in this format. –  Joachim Sauer Aug 2 '11 at 14:25
    
Sidenote: I don't think the SO technology is inadequate. It's great. But only for what it's designed to do (and possibly a few more side cases beside that). Being the only place for a thriving community to meet and interact, is not what it's designed to do. –  Joachim Sauer Aug 2 '11 at 14:29
    
@Joachim It's not being the community, for that there is Rbloggers, mailing lists and so on. It's about being a community where the QA format doesn't seem to fit into the rules of SO. What the R community deems a good and informative question, is closed and locked. So the R community basically wants to be able to decide for themselves what is a good question and what's not. –  Joris Meys Aug 2 '11 at 15:01
    
Keep in mind that the SE platform isn't precisely "bad for discussions". It's more that it actively opposes them. The SE platform was designed not only to be good at its intended purpose, but to discourage off-purpose use. That goes hand-in-hand with the rules against certain kinds of questions, even in cases where the questions are otherwise very good. I'm mostly involved with a fairly niche language, and our community uses mailing lists, IRC, and Reddit, among other places, not just SO. –  McCannot Aug 2 '11 at 16:08
1  
I'm actually partly convinced that a TeX-like solution would be appropriate, given that a big part of the R community doesn't come from the traditional SO-background. –  Joachim Sauer Aug 2 '11 at 16:11
add comment

There many uses of R to just get something done once (data analysis), and where the user is not specifically trying to program, so there are plenty of potential questions that would seem not to be a fit for StackOverflow. If nothing else, consider the use of R through GUIs (R Commander, Deducer, etc.).

I actually wish we could just ask these questions on CrossValidated, but the people there seem to have dug in their heels and want to push R-specific questions off the site.

Given both of these facts, there does seem to be a case for a new site, but...

Rather than finely chopping up the StackExchange network so much (and causing related problems of difficulty knowing where to ask questions, migrations, and dilution of reputation across similar sites), if there is going to be a separate site for using a statistics program, there should probably be a site for using all statistics programs.

  • This would avoid any eventual creation SPSS.stackexchange.com, SAS.stackexchange.com, Stata.stackexchange.com, etc.

  • This would be a pretty good match to the design of StackOverflow. There are many people who know how to use multiple stats programs, but may not ever bother to create an account on (or even visit) an R-specific site or a site for any specific program. CrossValidated really is the best place to find those people, but if the kind of questions we are talking about are kept out of that site...well...People might be more willing to sign up for a single other site to do with statistics.

  • This would also avoid Bill The Lizard's concern of "a tiny number of potential questions".


Edit: Thanks to Cody Gray for pointing out that on some parts of CrossValidated it seems their policy is to allow any questions on tools generally used by statisticians, in the vein of StackOverflow allowing questions on tools generally used by programmers. Unfortunately, there is disagreement and confusion about this. It seems that was the original policy, and some people agree with it, but others have caused it drift away from that openness toward setting a higher bar on the requirement that questions asked there require "statistical expertise" to answer. If that policy does not change back, it leaves a gap in StackExchange being able to handle the kinds of questions discussed by the OP here. What standard should be applied in deciding whether a question requires sufficient 'statistical expertise' to be posted to CV?

share|improve this answer
1  
It really seems like "using statistical programs to analyze statistical data" should fall within the purview of a statistics Q&A site... I mean, "programming tools commonly used by programmers" are on topic for Stack Overflow. –  Cody Gray Jul 22 '13 at 13:16
    
You would think so (+1 for sanity), but over on CV they seem dead-set against questions about tools. :S I wish I could explain to them that they could just filter out questions tagged with only a stats program if they really want to...or even have a tag specifically for stats program interface and filter out based on that. –  A.M. Jul 22 '13 at 13:23
    
Hmm, I've never even really been to the site before, but I just pulled up their Meta and immediately found this. I'm sure you know more about the community than I do, but you're not accidentally mistaking the policy against recommendation questions for one against all software questions, are you? Because Andre's answer pointing to the [software] tag makes me think that questions about using software for statistical or data analysis purposes are OK. –  Cody Gray Jul 22 '13 at 13:54
    
Naturally, they want to avoid programming questions, but those already have a home on Stack Overflow. I also found this question, which seems to confirm this view. Like you said, the people who don't care can ignore the questions. Works fine for me, I ignore R questions on SO. :-) –  Cody Gray Jul 22 '13 at 13:55
    
@CodyGray Thanks for the links. What's written there seems reasonable. Unfortunately, I have personally posted a question where they took a very wide interpretation of what "programming" is and a very narrow interpretation of what "to do with statistics" is. –  A.M. Jul 22 '13 at 15:25
    
Q on CV migrated to SO (and later marked as closed): stats.stackexchange.com/questions/62866/… ...and then closed as off-topic on SO too: stackoverflow.com/questions/17373838/… –  A.M. Jul 22 '13 at 15:26
    
You can see the CV migraters' reasoning here: meta.stats.stackexchange.com/questions/1640/… They even admit the question has zero to do with programming, but went to StackOverflow just because that's the only other place where a community of R users exists. That sounds like an endorsement for a new site. –  A.M. Jul 22 '13 at 15:36
    
Alternately, the non-programming uses of statistics programs could be hosted on SuperUser. I actually suggested that for that question, since it was clearly a better fit than StackOverflow. I expected to hear crickets if it went there, though. CV is the best choice, followed by a new site with a focus on stats-programs, followed by allowing use questions on SO without them actually being programming (something I don't think SO should have to accept). –  A.M. Jul 22 '13 at 15:37
    
For R, the answer may well be something non-SO/SE for the corner cases. Certainly SAS has its own user community (communities.sas.com) and I'd direct someone there who doesn't really fit the SO or CV models. I would say that CV probably should be more permissive of 'tools' questions, but I'm not a CV user so I am not going to tell them what to do. –  Joe Jul 22 '13 at 16:14
add comment

You must log in to answer this question.

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