9
$\begingroup$

This is a provocative question: should we send "How do I do this in R/SAS/Python/Stata/SPSS" questions to the DataScience, now that it's open?

Data cleaning is never a programming question, as it rarely is generalizable. SAS or Stata aren't really programming languages in the strict sense of that word, the way Python or C++ are.

So may be in addition to the existing "close -> off-topic -> another site -> StackExchange", we should also have "close -> off-topic -> another site -> DataScience". (May be at least some folks there would become aware of CV, as well. Most of what I do on DataScience is flag statistical questions to be migrated to CV.

$\endgroup$
  • 5
    $\begingroup$ I don't think beta sites can be listed in the flag/close dialog as migration options; I think one has to suggest that destination in an other (needs ♦ moderator attention) flag. $\endgroup$ – Nick Stauner Aug 14 '14 at 23:58
  • 2
    $\begingroup$ Some of those kinds of questions might be a better fit for SuperUser, perhaps. $\endgroup$ – Glen_b -Reinstate Monica Aug 15 '14 at 2:34
  • 2
    $\begingroup$ While whuber gave an extended, philosophical, and probably opinionated answer (I am not convinced that DS is bound to fail; I am only convinced that they will have to reinvent statistics on statistics questions, reinvent parts of StackExchange as far as the programming side goes, and reinvent parts of SuperUser as far as file management goes... but the same must have been said about CV at the time that half of it is mathematics, and the other half is programming)... anyway if what Nick Stauner say is technically true, I'd be happy to accept that as the answer, at least for the time being. $\endgroup$ – StasK Aug 19 '14 at 13:34
  • 1
    $\begingroup$ The first step should always be to remove the code from the question and leave it here. Or perhaps split the question and move the coding part to SO. The DS page should not have pure coding questions. DS is more about Machine Leaning, business cases for hypothesis test and general questions about coding frameworks. Pure coding questions should not be on DS just like how pure R questions should not be here even though they are likely to get answers. In fact R questions are likely better off here since most data scientists use Python not R. $\endgroup$ – Keith Aug 21 '14 at 18:27
6
$\begingroup$

My comment is true: beta sites can't be listed in the flag/close dialog as migration options; one has to suggest that destination in an other (needs ♦ moderator attention) flag. Here's a quote from Tim Post on Meta Programmers (this was as also quoted partially here on Meta Stack Exchange):

I'm Tim Post, a community manager for Stack Exchange. I'd like to take a few moments to answer this, and explain a bit about migration paths.

Migration paths between sites are something that we're very cautious to establish, we put them in place only when we're certain that the conduit will:

  • Help ease confusion for users that ask a question on a site that overlaps with another, ensuring a great question is sent where it will receive the best answers without much additional effort on the part of the question author,
  • Be largely beneficial to the target site,
  • Not become a party to unresolved topic disputes on the originating site, or between the originating and target site...

We don't establish migration paths to a site that is still in its beta period. We don't want to send content from one site to another unless we're certain that the target site is a permanent fixture. There are exceptions to this; moderators can migrate relatively new questions to sites not in the current migration path, but this ability should be used sparingly. A question should be of much better than average quality, and all hope of it being on topic for the current site should be exhausted prior to going around the established routes.

When [a site] comes out of beta, this is something that we can revisit. However, for now, it's simply not possible to add a migration path [to a beta site]. [Bolding added for emphasis.]

$\endgroup$
  • 1
    $\begingroup$ This doesn't preclude migrating to SO, which I think makes sense for at least R/Python questions (not as sure for SAS/Stata/SPSS). $\endgroup$ – Max Ghenis Apr 12 '15 at 1:12
19
$\begingroup$

I view data cleaning as on-topic here: it is a fundamental, if somewhat unglamorous (and underappreciated) aspect of data analysis. But, in keeping with this site's approach and philosophy, we would want to

  • welcome answers that focus on principles and generalities which would apply regardless of one's computing platform, and

  • migrate any threads that focus exclusively on which commands to issue or buttons to push in one particular platform.

Migration to DS is always possible: when voting to close as off-topic you could include a comment with this suggestion; moderators (who can designate any SE site as the target) can then act on that comment. I have migrated exactly one question to DS since it opened, so I don't think it's an act that will happen so frequently that it needs its own option in the list of closing reasons.

BTW, given that DS looks almost certain to fail, any question that might have any value here, no matter how peripheral, probably should stay here. I would advocate migrating only questions that are very clearly not on topic.

$\endgroup$
  • 6
    $\begingroup$ +1. Sending questions to Data Science will be sending them to oblivion if nobody there is likely to answer. I see no [stata] tag on Data Science, for example. Also, remember that there is surely no religious or practical objection to recommending a forum outside SE whenever that is likely to provide more support (e.g. Statalist). $\endgroup$ – Nick Cox Aug 15 '14 at 9:59
  • 1
    $\begingroup$ "given that DS looks almost certain to fail" Why would you say that? $\endgroup$ – jona Aug 17 '14 at 16:26
  • 1
    $\begingroup$ @Jona None of the metrics SE uses to evaluate sites are in the "good" or "excellent" range for DS and there are no signs they are getting any better. $\endgroup$ – whuber Aug 18 '14 at 13:52
  • 2
    $\begingroup$ @jona Info from area51.stackexchange.com/proposals/55053/data-science $\endgroup$ – Affine Aug 19 '14 at 4:20
  • 1
    $\begingroup$ Are CrossValidated statistics much better? I think whuber mentioned (somewhere else) that we are below one answer per question, I think, although we should be good on the number of high rep users (given that the site has been around for a number of years). $\endgroup$ – StasK Aug 19 '14 at 13:29
  • 1
    $\begingroup$ @StasK It would not be appropriate to compare the stats of a middle-aged site like ours to a beta. However, our beta was far stronger and graduated after several months. At the same stage the DS beta is languishing and gets little traffic. For instance, at three months we were getting 9 Q's/day, 20 A's/day, and over 1000 visits/day, while for DS the figures are 2.6, 7.2, and 190. $\endgroup$ – whuber Aug 19 '14 at 17:10
  • 4
    $\begingroup$ Most of the questions I see on DataScience would do perfectly well either on SO or CV. $\endgroup$ – shadowtalker Aug 19 '14 at 19:43
  • 5
    $\begingroup$ @ssde That's an important observation. I think it provides a plausible explanation for why DS is struggling: it hasn't been able to carve out a clear niche for itself. I actually do keep DS in mind when making migration decisions, but in almost all cases if a question is off-topic here it's because it is primarily of a programming nature where it belongs on SO rather than DS. That leaves almost nothing that truly qualifies to be sent to DS. $\endgroup$ – whuber Aug 19 '14 at 20:23
  • 1
    $\begingroup$ @StasK - the comparable beta stats for our site is here. The data dump says we have a current answer rate of 1.1 - for our beta it was 3. Looking at the past few weeks though (which I believe is slow for summer) we now get about 40 to 80 questions per day. $\endgroup$ – Andy W Aug 21 '14 at 20:27

You must log in to answer this question.

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