It's inevitable that we will get lots R questions in this stack exchange. Should we:

  1. Bounce them all to SO
  2. Answer all questions, even when it's clearly programming and not statistics.
  3. Answer the question unless it clearly has no statistical content.

I would vote for 3.

This should probably be expanded to include questions that cover any tool that has a substantial programming aspect: Excel, SAS, Incanter, ect, ect. –  Sharpie Jul 19 '10 at 21:53
I vote for 3, too!!! –  Vivi Jul 19 '10 at 22:51
Does the answer to this question (even though it is older) supersede…? –  rpierce Aug 10 '10 at 0:32
No we should follow the thinking in… The FAQs have been updated to reflect this change. –  csgillespie Aug 11 '10 at 13:15
add comment

6 Answers

up vote 19 down vote accepted

This is something that we deal with over at BioStar pretty regularly. Based on that experience, I'd argue that number 3 is the right approach, for the following reasons

  • There's going to be a fair amount of overlap between responders on the two sites
  • Bouncing people around will just lead an already confused person to become more frustrated. We want to be helping.

Of course, blatantly off-topic questions should should still be nixed. For borderline questions, I'd suggest that we give them a gentle nudge towards SO if they don't get answers here. Just suggest that they may get better responses there, and most people will be happy to make the jump over to the appropriate forum.

add comment

I also vote on 3. On the other hand, this site is finally the place which can accept purely statistical R questions from SO.

add comment

I'll say 3 as well, but there should be a very liberal criteria for determining whether a question has statistical content. For example, a seemingly purely programing based question (e.g. how to work with and store data in sparse matrices) should not be bounced to SO so long as the question asker specify that it is their ultimate goal to perform a specific statistical procedure on that data. There are many cases where it will be better to ask that type of question to people familiar with the procedure (and the available methods for conducting that procedure) rather than to a general programing forum. That is, a general programming forum may technically solve a problem but leave a dataset in a state that is not readily amenable to analysis.

Also, some CV users might want to migrate questions to SO based on elitist opinions of what questions require "statistical expertise to understand or answer". It seems important to be clear that any amount of statistical content is likely to render a question a better fit for CV than SO. I would wager that almost all people who deal with statistics also understand enough programming to handle statistics problems, but only a small amount of programmers happen to have experience with statistics. –  A.M. Jul 2 '13 at 3:19
@A.M.: I wouldn't bet that people who deal with stats understand enough programming to handle statistics problems. I've known some very savvy statistical thinkers who have never even so much as touched a punch card system let alone a modern programming environment. The two topics and modes of thinking are somewhat compatible, but the knowledge of one does not produce the other except under special circumstances. –  rpierce Jul 3 '13 at 3:03
As an aside, these old questions have been considered irrelevant for the ongoing community consensus by those who are of high reputation. So there is more to be gained by raising these issues on contemporary and active meta questions. –  rpierce Jul 3 '13 at 3:04
I suppose this is a case of "meta is different", but it doesn't quite follow the StackExchange M.O. for old questions to be abandoned and the same issues brought up in new questions (in other words, duplicates). ;) I am reluctant to raise a new one. I will try to find the newest question(s) (but, by the way, the newer ones I have found tend not to have as many votes). –  A.M. Jul 3 '13 at 3:15
I think it is more a case of the users who ended up with higher rep didn't like what came out of the old votes and to support their position claimed that this was the new standard and that somehow the old questions applied to the CV of the era in which they were asked. I think they are wrong, but to prove my point I added the summary meta thread -… - and I'll just keep banging that drum to keep people honest and from moving the yardstick. –  rpierce Jul 3 '13 at 10:49
add comment

Another vote for 3.

add comment

I would go for choice 3. Something along the lines of bouncing to Stack Overflow if the question is purely about implementation. Such as:

  • How do I do x in R?

  • Function Y/Script Z is not working, Help!

But if the question has to do with understanding some theoretical aspect of statistics and then carrying those concepts forward into implementation using a tool, R or otherwise, then I would vote that Stats should have a home for it.


I have been tagging questions that may be too heavy on the programming as "possibly-off-topic":

This way, we can easily locate them once we come to a consensus on this issue.

Retracted- see here.

If you find a question that you think is off-topic, link it to the appropriate discussion on meta so that it may be tracked. If there is no discussion on Meta yet, consider startin one!

add comment

I agree with 3 too, although as ever it'll depend on the contents of the question.

add comment

You must log in to answer this question.

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