5
$\begingroup$

Below question has been closed on Statistical Analysis web site. Is this really not a question for this site? I guess it the same as we ask Visual Studio question on stackoverflow, windows/mac questions on serverfault.

Access tables created in SAS Enterprise Guide Client into SAS Enterprise Miner Client?

Or if I am wrong, do we need a separate web site for such question?

| |
$\endgroup$
8
$\begingroup$

I don't think that questions that are strictly nuts-and-bolts software operation belong here. As I've ranted about before, that's how things are done on the GIS StackExchange, and I think it's very detrimental - the vast majority of questions are technical questions that would be better answered by asking on StackOverflow, RTFM, checking with the software company, etc. It's correspondingly very difficult to find questions that actually address interesting conceptual challenges. Bluntly put, it's boring.

To the argument that there are plenty of R questions of a similar flavor, I suggest that many of those questions should also be booted over to StackOverflow (including the one Srikant mentions, and this recent one about plotting). We certainly have to address statistical software if we're going to talk about statistical analysis, but in order to attract and retain expert users, we need to have questions that expert users are going to be interested in on the front page. If it's 20 questions about how to read data into Software X to one question about actual analysis, they aren't going to stick around. And if those who do stick around have to resort to using tags to sort through the data import questions, I suspect that we'll get many fewer eyes on each question as people select out only those questions that they "know" will be of interest.

Conversely, the StackOverflow R community has a wealth of questions that have already been answered, and a stable body of experts who frequently answer questions within minutes. The same isn't true of SAS, but SAS questions do get answered there and there are plenty of other community resources for SAS, STATA, etc. (and, as any programmer can tell you, community is also a feature).

| |
$\endgroup$
  • 1
    $\begingroup$ @Matt I am sure many experts would agree that data visualizations are as important as statistical analysis. See today's post on line plots by Andrew Gelman as an example. By the way, while I myself share your opinion that they are boring, do also note that what may be boring to one person may be exciting to another. In addition, generating good plots is hard and requires several years of practice and hence needs expert answers. $\endgroup$ – svadali Sep 2 '10 at 14:27
  • 2
    $\begingroup$ I agree, and I fully support the inclusion of data visualization at this site. I don't agree, however, that every question relating to the technical production of plots is appropriate for this site. If the questions relate to the conceptual and aesthetic aspects of a plot, then I think that's fine; if it relates strictly to the programming required to produce said plot, I don't think that's appropriate. If someone asked how to reproduce a plot in Processing, for example, I think it would be more clear that that would be better off at SO. $\endgroup$ – Matt Parker Sep 2 '10 at 15:07
  • 1
    $\begingroup$ And with regards to things being boring: you're right, of course. But the reason we have multiple Stack sites is so that people can maximize the amount of interesting stuff they see. I'm trying to ensure that experts in statistical analysis, machine learning, and data visualization see interesting things when they come here - not data IO questions. $\endgroup$ – Matt Parker Sep 2 '10 at 15:10
  • 3
    $\begingroup$ Let me put on my marketing hat for a second and address the issue from that perspective. The site's customers are experts and novices in data analysis (broadly speaking). Clearly we need both kinds for a site such as ours to thrive. So, what are the needs of our customers. They can vary a lot but I would classify them in the following broad categories: (a) Ask/answer theoretical aspects of data analysis and (b) Ask/answer practical questions of data analysis. $\endgroup$ – svadali Sep 2 '10 at 15:15
  • 5
    $\begingroup$ A practical statistician would perform the following tasks: 1. Exploratory data analysis, 2. Analyze the data using several models, 3. Evaluate model fit and 4. Generate plots to communicate the results of the analysis to a wider audience. We should be a 'one-stop-shop' for the practical statistician. From a marketing and convenience perspective, it does not make sense to tell folks that for A you go to staoverflow, for B you go to stats.stack etc. I will stop here and let others chime in with their thoughts. $\endgroup$ – svadali Sep 2 '10 at 15:17
  • 2
    $\begingroup$ I agree with you in theory, but let me list some technologies I work with as a practical statistician: R, SAS, Stata, Python, Git, Access, Excel, Word, PowerPoint, Visual Basic, UCInet, ArcGIS, ODBC, Windows XP, Ubuntu, Tinn-R, Notepad, FTPzilla, XML, CSV, TSV, JSON, Internet Explorer (6!)... I use all of those things routinely in analysis. Should we answer questions about all of that stuff here? No; I'm being over the top. From (my) marketing and convenience perspective, it doesn't make sense to make people wade through all of that stuff to browse the things in which they're interested. $\endgroup$ – Matt Parker Sep 2 '10 at 16:00
  • 1
    $\begingroup$ We're never going to have a definitive rule, and in practice I favor being pretty lenient with questions like the ones mentioned above. But when we're setting policy, I think we need to keep it close to the stats. I'm actually very happy with the way things are right now - I think we've hit the right balance naturally - so I'm wary of any policy changes that might disturb that. I'm now going to follow your wise example and see what everyone else has to say. $\endgroup$ – Matt Parker Sep 2 '10 at 16:06
  • 1
    $\begingroup$ I disagree that there is any "dilution"; we have tags, we can filter out what we don't want to see. On the other hand, what we really need is some usefulness-mass that will attract googlers; this is something easier to build from technicalities. Finally we need "easy" questions that will enroll newcomers in the reputation marathon. Without those, we will have a private chat place rather than SOFU-class site. $\endgroup$ – user88 Oct 28 '10 at 9:00
7
$\begingroup$

I think it belongs on the site. As another example of a question similar to the SAS question, consider this question on how to use R to draw a particular type of plot.

Both types of questions deal with how to use the software (R, SAS) to perform a type of operation (draw plot, import data) that is needed as part of a data analysis project.

I voted to re-open.

| |
$\endgroup$
  • $\begingroup$ +1 from me..... $\endgroup$ – csgillespie Sep 2 '10 at 13:18
  • $\begingroup$ I tend to think this type of question should be allowed. So I support re-opening. $\endgroup$ – Rob Hyndman Sep 2 '10 at 13:31
  • 1
    $\begingroup$ I agree w.r.t. to the SAS question, but not for this question: stats.stackexchange.com/questions/2282/…. I am fine with questions about how to use tools in general, but I don't think that we should have major discussions about the tools themselves. $\endgroup$ – Shane Sep 2 '10 at 14:35
  • $\begingroup$ BTW. That's three votes from moderators, so I will reopen the question. The community can vote to close in a normal fashion on this question going forward. $\endgroup$ – Shane Sep 2 '10 at 14:35
  • 2
    $\begingroup$ One final comment: I think that this is a very bad question. It doesn't give any background or provide any reason for people to consider it outside of this small problem. I much prefer having questions that will be of more general interest. $\endgroup$ – Shane Sep 2 '10 at 14:38
5
$\begingroup$

I voted to close the question the OP is referring to, and I think it is innapropriate for this forum. While people on this forum could potentially answer the question, I think it is more appropriate to refer people to expert bodies specializing in those particular software packages. People are not only more likely to get a quick response, but should search through old posts on those sites to see if the question they have has already been answered.

I fail to see how that question on importing data between different SAS products has anything to do with statistical analysis. If it wasn't SAS, and say a person was asking how to conduct a SQL query in Microsoft Access would the question be considered appropriate? The difference between that question and asking about visual studio on stack overflow is that statistical software has a broader range than statistical analysis, and that question falls squarely into what I would consider "data management". Although as Srikant points out this is an essential tool to conduct analysis, I feel it is far enough removed from anything resembling statistical interpretation that it dilutes what I believe the intentions of the sight are (to promote discussion and learning of statistics).

I would be alittle more sympathetic except for the fact that many of these software packages have user groups that individuals would be better addressing their questions towards (and searching their archives of past questions). I don't know for SAS, but if the software was SPSS I could refer the individual to several user groups where SPSS experts and even representatives of the software post responses to questions. I know Stata and R have similar listserves as well. Why should this site be redundant with these groups at the expense of this sites members time?

| |
$\endgroup$
  • 4
    $\begingroup$ Maybe we could work on putting together a pretty comprehensive list of such groups to add to the FAQ. Then, whenever such software questions came up, we could just point askers to that list. $\endgroup$ – Matt Parker Sep 3 '10 at 13:49
  • 2
    $\begingroup$ I was thinking about posting a community wiki question on the main forum where people list resources for each of these software packages. I'm not sure whether it should be posted in questions, on tag wiki's, or in a FAQ (or all three), but I agree the site should have such a list to direct people towards for different packages. $\endgroup$ – Andy W Sep 3 '10 at 15:10
  • 2
    $\begingroup$ This is an interesting debate and I find it more of a debate about commercial vs open source software. There are ton of R-specific questions on this site already. I can't claim to have done an exhaustive search but I cannot find any R-specific questions that have any comments regarding "this should be on another forum". And there are certainly plenty of R forums. That there is a double standard is understandable, but let's call a spade a spade. Most people on this site don't mind the R and Python related questions because those are open source tools. $\endgroup$ – Josh Hemann Sep 5 '10 at 2:51
  • $\begingroup$ @Josh, sorry for the late response. I understand where you are coming from, although in my opinion whether the software is SAS or R makes no difference to me. If the question was "How do you import a csv file into R?" I would have thought it was inappropriate as well (although probably over 100 people on the forum could answer the question in less than a minute!). I would think of the divide more so in terms of what people are trying to do with the software, as opposed to what the software is. R probably has no equivalent to the SAS products the original question asked about. $\endgroup$ – Andy W Sep 7 '10 at 3:14
  • $\begingroup$ Sorry Andy, I was not trying to single you out, others have had similar views. There is a post today about an error running an ANOVA in R and which package to use to get a sphericity. A reasonable question, and no one has said anything about it not being appropriate, but to me it fails the criteria listed in this debate for what is a valid question. If the exact same question was about an error running ANOVA in SAS I think there would be numerous comments about "post somewhere else". Hopefully SA keeps focus on analysis but it is impossible to detach software from the discussion. $\endgroup$ – Josh Hemann Sep 7 '10 at 16:04

You must log in to answer this question.

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