23
$\begingroup$

Greetings from your friendly neighbors who frequent the tag over at StackOverflow!

We are all familiar with the fact that questions involving R will often straddle the two sites, and so a certain amount of migration between the two is to be expected.

However, a few of us over at SO have noticed that recently it seems as though R questions are being migrated to SO regardless of their quality. The latest example was this question. (A few more examples of what I would consider low quality here, here, and here.) That question is both rather unclear and certainly does not display any research effort on the part of the asker.

I would just like to remind everyone, whenever possible: never migrate crap!

Instead, it may be better to try to encourage the OP to improve the question on your site, and if they do, then feel free to migrate it. Otherwise, just close it.

My question is why are these low quality questions be migrated at all?

(None of this is meant to imply that low quality questions are never migrated in the other direction, from SO to CV. Please feel free to pop into the R chat room at SO to give us feedback on any low quality migrations. I simply felt like I'd seen more low quality migrations than usual recently, so I thought I'd mention it, just as a reminder.)

$\endgroup$
  • 1
    $\begingroup$ +1 It seems to me that many of these migrated questions fall in the category "data visualisation", which is clearly in scope for CrossValidated. Here are some more examples: stackoverflow.com/q/11610377/602276, stackoverflow.com/q/11544051/602276, stackoverflow.com/q/11449584/602276 $\endgroup$ – Andrie Jul 26 '12 at 17:58
  • 4
    $\begingroup$ @Andrie Those particular questions aren't really in scope for CV, because they're strictly questions about implementation of visualizations in a particular language. To paraphrase the FAQ: "If your question requires statistical expertise, ask here - otherwise, ask elsewhere." $\endgroup$ – Matt Parker Jul 26 '12 at 18:15
  • $\begingroup$ @MattParker That makes sense, but ever since I was lightly wrapped over the knuckles for daring to suggest that to a CV question, I have tried to keep myself to myself. $\endgroup$ – Andrie Jul 26 '12 at 18:18
  • 7
    $\begingroup$ Crap is in the eye of the beholder. I would take low quality to typically mean that the question is ill-phrased or hard to answer because a lack of critical detail (or just very poorly written/formatted). I wouldn't take trivial to answer (or RTFM) questions as crap. Perhaps you should post an answer to dictate examples of what should and should not be migrated. $\endgroup$ – Andy W Jul 26 '12 at 18:27
  • 4
    $\begingroup$ One of the things I have found difficult in deciding whether it should be closed or migrated are questions I suspect are R questions already asked and answered on SO (no point in migrating a duplicate right?) But is it reasonable to expect the CV people to find the duplicate on SO before determining if it should be migrated or just closed? $\endgroup$ – Andy W Jul 26 '12 at 18:28
  • 2
    $\begingroup$ @AndyW I'm less concerned with duplicates being migrated than with questions that are generally low quality. For instance, questions where the OP says something like "I want to do X. How do I do that in R?" without providing any evidence that they have tried something themselves. $\endgroup$ – joran Jul 26 '12 at 18:31
  • 3
    $\begingroup$ @Andrie Hm! If it's any consolation, I completely agree with your suggestion there (and I think it's completely in concordance with our "statistical expertise" guideline). $\endgroup$ – Matt Parker Jul 26 '12 at 18:32
  • $\begingroup$ @MattParker Then all is forgiven! ;-) $\endgroup$ – Andrie Jul 26 '12 at 18:33
24
$\begingroup$

I'm quite sympathetic--as a mod, I don't like bad questions from other sites landing here, either. I greatly appreciate your reaching out to us with your concern (and not just swearing under your breath at the boneheads over on CV... :-). Your reminder will cause me to pause a little longer in the future before hitting that "migrate to SO" button. But please forgive me for replying with some questions of my own, because it seems like some implicit assumptions are operating here and, in the interest of good mutual understanding, we ought to unveil them and briefly examine their implications.

  1. When a question is off topic here, then why should we make an effort to improve it before closing or migration? Nevertheless we usually do, but at the point it becomes clear that a question belongs on another site, then the norms of the target site apply. Wouldn't it then be presumptuous of us to continue to work on the question?

  2. Following up on that one, it seems to me that if we simply close an off topic question (but not brusquely: of course we would give a reason and it would be that the question is off topic here and it might include some additional constructive comments), then any sentient asker who is still breathing will just head over to another site and re-ask their question there, anyway. So you guys get to start over without the benefit of any comments or suggestions we would have made along the way.

  3. The main reason certain questions are being characterized as "low quality" appears to be that they are RTFMs. This has been debated on SO meta. My read is that (a) these are acceptable questions unless (b) the asker is a serial abuser. Are you then suggesting that CV community members should research the asker's history on the target site before voting to migrate? This seems like it would be asking a little much, and most of us don't have adequate privileges to do this anyway.

  4. R documentation in particular is notoriously bad: it is not a trivial matter to find the right parameter in the right function in the right package to do what you want and even then the manual pages may be opaque (to be generous about it). I agree that providing evidence of research makes for a great question. But it seems like a lot to insist that all questions must document a fruitless search in order to be asked.

When I look at the four examples offered I agree that none of them is a great question. Yet none of them are on topic here; all are on topic on SO; and three of them are clearly formulated and have answers demonstrating these aren't completely trivial to solve. That "latest example" is quite clear, IMHO, and the answer requires not only research of the correct parameter, but also knowledge of string handling in R (via sprintf) to solve. Trivial for you and anyone reasonably familiar with R to answer; certainly. So it seems the main objection here is that "easy to answer" = "crap." Fair enough, if that's the norm on SO (but I suspect not), yet it doesn't seem possible to turn that sentiment into actionable criteria that this community can be reasonably expected to follow.


By the same lights, although (IMHO) the majority of questions migrated to CV are garbage, we understand that it would be unfair and unreasonable to ask the SO community or moderators to vet statistical questions for clarity or meaningfulness: you cannot be expected to have the knowledge or interest to do this. (We close a lot of them right away.) But please keep sending them: helping people improve statistical questions is something we are good at and we want to be the ones deciding which questions are worth keeping.

$\endgroup$
  • 4
    $\begingroup$ big +1, I wholeheartedly agree w/ this. Moreover, I want to say that CV & SO are question and answer sites. The reason people are asking is because they don't know. Certainly there are questions that are garbage & don't show any effort on the part of the OP, these are commented, ignored, downvoted, edited, and/or closed. I've been as frustrated as anyone. Nonetheless, this is a Q&A site, and our 1st obligation is to give the benefit of the doubt & try to work w/ the person to improve (hence the "serial abuser" part). $\endgroup$ – gung Jul 26 '12 at 20:01
  • 5
    $\begingroup$ SE sites are not, IMHO, primarily about providing an intellectual challenge for high-rep users. If your attitude is that 'any question that's easy for me to answer is crap', then maybe public Q&A sites aren't for you. $\endgroup$ – gung Jul 26 '12 at 20:01
  • 4
    $\begingroup$ OK, @gung, but maybe I should have made it clearer that I don't think this attitude applies to joran. It just seems that the easy=crap equation may be a logical consequence of what we are being asked to do, and in light of that we should be thinking harder about what our criteria for closure and migration should be, as chl is asking in his reply. $\endgroup$ – whuber Jul 26 '12 at 20:05
  • 3
    $\begingroup$ Thanks for the detailed responses. I'm tied up this afternoon with a small infant, but based on this and @chl's response I think it's a good thing I raised this issue as I suspect it will help highlight some differences between the two sites. Once the little one is in bed tonight, I'll post a more thoughtful reply. $\endgroup$ – joran Jul 26 '12 at 20:51
  • 2
    $\begingroup$ I don't get the slight at R's documentation, "notoriously bad" is the kind of thing someone would say when they hear bad press but don't check for themselves, though that's clearly not quite the case here. R's documentation is amazingly good. $\endgroup$ – mdsumner Jul 27 '12 at 1:54
  • $\begingroup$ As far as I know, the usual process in case of doubt is to ping the target site's mods in TL. That might not work well here as your target is a particular tag on SO, but I guess you guys could easily establish a chat room for migration purposes. I would expect you to come to mutual understanding of standards quite quickly, and you can communicate before migration (less user frustration). $\endgroup$ – Raphael Jul 27 '12 at 9:45
  • 6
    $\begingroup$ @md I would love to know what documentation you are reading. Most of the stuff I find consists of these alphabetized abstract lists of functions and classes offering no significant examples of use, no theoretical explanations (apart from literature references), and amazingly inconsistent conventions for passing arguments and structuring output. This is a natural consequence of software organically built by a large disparate user community, so I'm not condemning it, but it's a huge stretch to call this stuff "amazingly good." $\endgroup$ – whuber Jul 27 '12 at 13:25
  • 2
    $\begingroup$ @whuber: I think it's very important to distinguish between R-base/recommended and user-contributed packages. R-base/recommended documentation is better than most of the commercial software I've used, and--while I agree that many user-contributed packages are very poorly documented--I don't think it's fair to broadly attribute that to "R". $\endgroup$ – Joshua Ulrich Jul 30 '12 at 21:52
13
$\begingroup$

This is a fair reminder and an interesting occasion to discuss how to handle such question. When the question has to do with statistical analysis or related techniques (including visualization when the purpose is to answer a statistical question)--as per our FAQ--it belongs here. Low quality questions are either closed or subjected to a request for clarification. This is an important aspect of Stack Exchange of giving the opportunity to the OP to clarify anything that might be misinterpreted, or to enhance the quality of his post to fit within SE guidelines.

The case of R questions has been debated more than once on our Meta, and the older thread probably started during the beta: How to answer R questions. Purely programming questions are generally sent over Stack Overflow. Low quality questions, however, should not. Apparently, the question that is referenced in your question is part of this category, according to your criteria which are certainly well founded, given your experience, and that of Andrie, on SO. It is, however, hard for some of us to help the OP reformulate his question so that it sounds acceptable to you, as a community. Asking SO moderators before migrating is one option and CV moderators do that on some occasions. Given the large part devoted to the use of R in statistics on this site, many questions are likely candidates for SO when they really don't have anything to do with statistics per se, but I think it would be hard to ask SO moderators each time.

I would love having a well defined set of criteria that a question must fullfill before it is considered as belonging on Stack Overflow and not any other site. But it is worth noting that qualifying a question as crappy is risky: After all, very simple but well posed question are eligible according to our common FAQ:

You should only ask practical, answerable questions based on actual problems that you face (...)

That being said, you and I know how it can be difficult to judge the quality of a question when we are not following a site on a very regular basis (this is my case as a user on SO, at least). It is even harder to decide whether to close a question as a duplicate, or simply as a badly worded question when it happens.

To sum up, I would be happy to have some concrete illustrations of

  • What is considered strictly OT on SO and should be closed immediately (possibly, with a sensible comment to accompany such action).
  • What is considered as a ready-to-migrate question, i.e. without any intervention from our part.

Finally, since you are active on both sites, what would you suggest we put in our FAQ to help users decide on the best site for their question, even if many new users probably don't visit our FAQ before asking their first question :(

$\endgroup$
7
$\begingroup$

I appreciate the thoughtful responses from both @chl and @whuber.

One thing I'd like to clear up right away is that the issue is most certainly not that we don't like simple, or trivial questions over at SO. For instance, take this question from just earlier today. That was extremely trivial to answer (and answered by one of the highest rep R tag SO users!), but I think everyone over at SO would be delighted if all our questions were like that one. The asker provided a clear, reproducible presentation of their data and code, and clearly described the problem.

I think there are two general categories of questions that some of us find disagreeable:

  1. "I have this data. I want a graph that looks like X, Y and Z. How do I do it?" with no code, and the data isn't provided using something reproducible, like dput.
  2. Just generally unanswerable, like one of my examples above:

    I have 2 variables that, when combined, uniquely identify a product (barcode id and catalog descriptions). Barcode numbers and catalog descriptions have drifted over time, but not at the same time (e.g. barcode value for one product changed in May, then catalog description slightly changed in December). I would like to create some kind of object to identify these drifts (e.g. when barcode value xxxxxxx, connect all catalog descriptions, then with all catalog descriptions, connect all barcode values, then continue until no more matches).

    Any idea how to implement this?

The problem with questions like in (1) is that they break the Q&A model. The questioner is no longer asking for help with a specific problem, they are asking someone to do something for them. This type of thing seems to be happening more and more regularly with data visualization questions. The problems with (2) should be pretty apparent.

As for solutions, I think both @chl and @whuber make some good points regarding the difficulty for CV users to magically fix questions before migrating them, and the fact that askers will always be able to simply re-ask a closed question elsewhere.

In an ideal world, I would like R questions that are migrated to SO to at least demonstrate some level of research effort. This could take many forms. Example code that illustrates the undesired outcome and data (ideally reproducible) would be best, of course. But really anything that reasonably illustrates what the asker has tried, and why it didn't work.

Returning to my first example above, in its original form:

I have plotted some numbers in a scatterplot in R and want to add dollar signs to the numbers. How would I do this?

I could quibble about just how clear this is: dollar signs where? The axis title? The axis tick labels? Or did you plot numbers as text in the scatterplot itself?

But if they had at least shown the code they'd tried up to that point, most of those questions would be moot. And even a few cursory comments along the lines of "I looked in ?plot and ?points but didn't find anything helpful" would be an improvement.

That said, I recognize that it may be difficult for your community to do all the work to improve questions prior to migrating. I'd settle for:

  • just not migrating the truly horrible questions (i.e. the obvious cases, like (2) above), and
  • when you do migrate a question that "needs work", send it over with a polite comment suggesting they edit their question following the guidelines here. I know it seems like a trivial thing, but if the question already has a comment along those lines, it's a lot easier for folks who are easily annoyed to simply ignore it, and it saves some work for those of us who want to be helpful.

More generally, and this is mostly an aside, I think there's sort of an Urban/Rural divide going on here between SO and CV. A lot of behaviors that can seem rude or abrupt in an urban setting are natural adaptations to coexisting with a much higher population density. I think that SO can seem a lot harsher about issues like this than on other smaller StackExchange sites, but it's not always because we're just big meanies over on SO. :)

$\endgroup$
  • 2
    $\begingroup$ (+1) Thanks for clarifying those points. As I am short of time right now, I will comment on later. I've created a chat room if anybody want to discuss in a more direct manner. $\endgroup$ – chl Jul 27 '12 at 7:23
  • 2
    $\begingroup$ We have a library of helpful responses, we could probably adapt one to this particular situation (it certainly occurs more than some of the other helpful responses listed so far). $\endgroup$ – Andy W Jul 27 '12 at 12:08
  • 2
    $\begingroup$ +1 This is great post: thank you for taking the time and effort to respond to our questions and concerns. The "urban/rural" metaphor seems to capture much of what's going on. It gives me some ideas about how to better evaluate questions we get here, not just those we intend to migrate. $\endgroup$ – whuber Jul 27 '12 at 13:29
  • $\begingroup$ @AndyW That's a good idea. It might also be possible to create a dedicated entry, and perhaps link to it from our FAQ. It would also be accessible from main sites (SO and CV). $\endgroup$ – chl Jul 27 '12 at 13:29

You must log in to answer this question.

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