# Defining clearer house rules on software-related questions

Given the various issues regarding handling of software-related questions (see here for one jumping-off point), can we curate a set of house rules that

1. encapsulates what Question the community would like to see remain on CrossValidated (CV)
2. provides explicit guidance for users and moderators alike when considering flagging for migration or migrating Questions to StackOverflow (SO)

I have made an initial attempt, see my Answer below, but as that is classed Community Wiki please feel free to make edits to it as well as adding additional Answers that might act as signposts for discussion or jumping-off points to discuss particular aspects of suggested wording.

Note that the base text was lifted from the What topics can I ask about here? page.

I decided to reproduce the entire document from the above link as it contains a lot of useful material that we should maintain.

• I wanted to reference a previous discussion which included the CV/SO division and migration: meta.stats.stackexchange.com/q/1335/5880 – Brian Diggs Sep 18 '13 at 15:15
• While this is a frequent topic of meta-discussion, how large of a problem is this generally? Read: Is the time spent rewriting the FAQ and essentially suggesting an algorithm to consider for each new post <= the time spent on the couple clicks it takes to migrate a question to SO? This feels like a leading question, but it's genuine, I don't have a good feel for the volume of "pure programming" questions here. – Fomite Sep 25 '13 at 21:05

New text begins below the line, other horizontal lines enclose regions where changes have been made

CrossValidated is for statisticians, data miners, and anyone else doing data analysis or interested in it as a discipline. If you have a question about

• statistical analysis - applied or theoretical
• designing experiments
• collecting data
• data mining
• machine learning
• visualizing data
• probability theory
• mathematical statistics
• statistical and data-driven computing

then you're in the right place. Anybody can ask a question, regardless of skills and experience, but some questions are still better than others. If you came here with a question to ask and are new to the site, please consult our thread on how to ask a good question.

Our community aims to create a lasting record of great solutions to questions. For more about this and guidance about how to provide your own great answers, please read How should questions be answered on Cross Validated? Providing references to peer-reviewed literature or links to on-line resources is warmly welcomed. You can also incorporate the work of others under fair use doctrine, which particularly means that you must attribute any text, images, or other material that is not originally yours.

[Edit: Converted the Homework link to SE markup for the Homework tag]

Homework questions are welcome. Please mark them with the tag. They get somewhat special treatment, because ultimately you benefit most by finding the solution yourself. The community will try to provide guidance, hints, and useful links.

For more help, see What types of questions should I avoid asking?

## Heavily modified!

• Reordered bullets, collected all software-related entries into a single section
• Under the programming sub-bullet I added a Note to clarify what we mean by programming. The other text in this bullet is a verbatim copy from the original
• Added several bullets under the general topic (which is also new) of Software-related Questions.

There are certain subjects that will probably get better responses on our sister sites. If your question is about

• Bugs in software, ask the people who produced the software.

Added: For add-on packages for languages such as R, written by third parties, identify the author or package maintainer so as to make your bug report in the appropriate place.

• Software-related questions Computers and statistical software are integral to the modern practice of statistics or data analysis. As such, questions with a significant software aspect are in general welcome on CV. However, the CV community has no desire to replace documentation or existing help email lists or discussion fora for statistical software. Consider whether a software-related question is suitable for CV via the following criteria:

• Software-related questions that are potentially related to the practice of statistics or data analysis, however well-disguised as a coding or software issue they may be. Such questions should be assumed on topic for CV unless another criterion would render them off topic. Such questions may need improvement to meet other quality criteria; assist the original poster to improve the question to draw out the statistical or data analytical issues by leaving comments.

• General questions about using statistical functions, software or add-on packages for languages such as R, especially those concerning basic usage, are off topic on CV and SO, particularly so when they concern topics covered in the introductory manuals, or function or package documentation. As such, these questions should not in general be migrated or flagged for migration to SO.

• Programming If the language is statistically oriented (such as R, SAS, Stata, SPSS, etc.), then decide based on the nature of your question: if it needs statistical expertise to understand or answer, ask it here; if it's about an algorithm, routine data processing, or details of the language, ask on Stack Overflow or refer to the collection of links to resources we maintain.

Note: With languages such as R or Python it can be difficult to distinguish between programming and simply using the software's functionality for data analysis. Questions about using higher-level statistical functions or add-on packages in such languages are not considered programming. Do not flag such questions for migration to SO.

• If statistical reasoning is required to answer the question, the question is on topic. This criterion has broad applicability, for example a question requiring help with debugging code producing patently incorrect results is considered on topic if statistical reasoning is needed to correct the code.

Questions about obtaining particular datasets are off-topic (they are too specialized). The GIS SE site welcomes inquiries about obtaining geographically related datasets.

Please note, however, that cross-posting is not encouraged on SE sites. Choose the best location to post your question. Later, if it proves better suited on another site, it can be migrated, along with any comments and answers. In particular, don't post a duplicate question on another SE site if your question attracts close votes; let the community moderators migrate the question for you. To aid this process, feel free to flag your own question for moderator attention.

If your question is not specifically on-topic for CrossValidated, it may be on topic for another Stack Exchange site. If no site currently exists that will accept your question, you may commit to or propose a new site at Area51, the place where new Stack Exchange communities are democratically created.

• The above is my view on what the house rules should be. I have however tried to incorporate the thoughts of other prominent Cross Validated users as expressed in other related threads. I appreciate that not everyone will agree with the viewpoint stated above. Rather than edit-war this Answer, let's discuss this via other Answers & comment threads and hopefully come to a consensus that is edited into this Wiki answer. – Gavin Simpson Sep 18 '13 at 5:07
• A few concrete examples might help explain the software-related sections. For example, CV won't explain how to read your data into R or SPSS, but will explain why the dendrograms produced by each package have a different axes (or wherever the line ends up being drawn) – Matt Krause Sep 18 '13 at 5:22
• I am afraid the part on software-related questions just adds more confusion. We have “Questions about using higher-level statistical functions or add-on packages in such languages are not considered programming” but then “General questions about using statistical functions, software or add-on packages for languages such as R, especially those concerning basic usage, are off topic.” It seems to suggest that we want to continue current practice, merely acknowledging that questions about usage aren't about programming. – Gala Sep 18 '13 at 6:28
• If that's what it is, it would be better to specify that those questions are off topic in the first paragraph as well and leave it at that. (I am not convinced the language about “general” or “basic” usage is very helpful either. A function is only “basic” once you know it and use it regularly, you wouldn't ask about it if it was easy for you.) – Gala Sep 18 '13 at 6:32
• Readers should be clear that the above is not exactly what is on the on-topic page of the help center. The "Note" under the "Programming" bullet point and the final bullet pointed section "Software-related questions" are suggested changes to CV's policy. I think it would be better to have the current text reproduced & then have the suggested changes as a distinct answer to make it easier for people to evaluate the status quo & possible changes. – gung - Reinstate Monica Sep 18 '13 at 13:23
• I agree with @gung: What we really need to know are what changes are being proposed to our FAQ. We can all read the FAQ as it stands now. Producing the entire new text asks us to work much harder than necessary to identify and react to the proposed changes. – whuber Sep 18 '13 at 13:38
• @gung Yes, I see that what is above could cause confusion. I genuinely wasn't sure how to best approach this - the Q&A format isn't a great fit for collaborative editing a document. I thought having an entire whole proposed text would be useful with other Answers curating specific elements that could be edit in. I have to mend a puncture just now but will edit the above to indicate where changes have been made. – Gavin Simpson Sep 18 '13 at 13:43
• Be well, @GavinSimpson. When you're ready, here is what I would suggest: in the question, add a horizontal line & then reproduce the text under the status quo. For your answer, place the changes you would like to see, possibly with a couple of notes about where they would fit. Then add a horizontal line & you can provide discussion / arguments in favor of your suggested changes. – gung - Reinstate Monica Sep 18 '13 at 13:47
• I agree with the sub-bullet points under "software-related questions" but fear that "potentially related to the practice of statistics" is so general that it's almost meaningless. Maybe a clause along the lines of "in an obvious way" should be added. But, at that point, I think it completely overlaps in content with the third sub-bullet point, which I think gets at the heart of it: if it requires statistical reasoning, then it's on-topic. The problem is that I think this is the criteria that most close-reviewers already use, which is inherently subjective and has lead us to where we are now. – Macro Sep 18 '13 at 13:59
• @gung I decide against importing the existing text - it is useless to scroll back an forth to compare and anyone can read the proposed version here side by side with the original using their window manager and browser tabs. I will make it clear that this is a proposed change though. – Gavin Simpson Sep 18 '13 at 16:15
• That seem clear enough, @GavinSimpson. – gung - Reinstate Monica Sep 18 '13 at 16:16
• @GaëlLaurans I have now edited the proposed text to include all software-related items under a single top-level bullet. I have pulled Programming into that section as it fits better there. I see this now after reading your comment; Thanks I also added a sentence to clarify the point of the Note; it is to help with guidelines on migrating questions to Stack Overflow as there is a lot of confusion on what should be migrated. – Gavin Simpson Sep 18 '13 at 16:24
• @Marco Thanks. I think that adding an "in an obvious way" clarifier misses the point that a number of users have made in comments under various Meta Qs, that the OP may not appreciate the statistical issues. These may not be obvious to all, but a knowledgeable user could draw out those issues in a comment or answer. That would be a net win. I am wary of adding what you suggest as that gives an easy route to flagging or close-voting. – Gavin Simpson Sep 18 '13 at 16:26
• @GaëlLaurans As regards the use of "basic", I don't equate that term exclusively with "easy", it is more nuanced than that. Perhaps "introductory" is what I meant? – Gavin Simpson Sep 18 '13 at 16:36
• I would note that the line between 'programming' and 'using functions' is rather blurred. Take this from wikipedia: "A computer program ... is a sequence of instructions, written to perform a specified task with a computer.". If someone wants to combine a couple of R (say) function-calls, by that definition, they have a program, and are thereby programming. If we're just going to shift the argument to 'what really constitutes programming' I don't think we've made things clearer, but perhaps even more murky. – Glen_b Sep 19 '13 at 23:44

All of the discussion reminds me of bureaucrats aiming to impugn a strict set of rules to a diverse set of circumstances. I'm skeptical such a list of rules can be derived or agreed upon. To hijack your question (sorry) I'd prefer to focus on actionable steps the community can take to better resolve such disagreements. I'll propose two;

• A sizable proportion of such disagreements are over ill-defined questions (oft by new users). We can encourage reviewers to be less hasty in voting for migration when such questions occur and encourage other actions such as editing (which can be more easily reversed).
• A forum for discussion on borderline questions (much like the one Gavin suggests the R community on stackoverflow has). Whuber states we migrate on average 2 questions a day; this is few enough discussion can be taken on an individual question basis when disagreements arise. This can be accommodated via chat.

Of course as best we can do is encourage such behavior (we can't force individual action). The biggest impediment at this stage IMO is getting adoption of the chat system. Fingers crossed, we may have enough community participation at this point (if we prod people enough) to establish regular participation.

I just don't think adding to the already long list of guidelines has any utility to the site. I'm skeptical the FAQ is an effective vehicle to encourage behavior for either newcomers (asking questions) or regular users (the ones voting to migrate).

• Being likened to a bureaucrat is like a nice to my heart ;-) Point taken; I would liken it to herding cats! As for having a better resolution mechanism is excellent and is needed as we can add a note to that effect in the proposed text. What you envisage is exactly how we curate Qs under the R tag on Stack Overflow. A link to the Q is posted in chat and an explanation as to why a close or migrate vote/flag is justified is given. That helps get eyes on a question thus expediting the Review process, but it also allows discussion if there is disagreement. – Gavin Simpson Sep 18 '13 at 16:32
• Some further background for interested readers: On Stack Overflow the R chat room has been made private (you need a certain rep to contribute or by invitation for valuable members of the community). This has helped weed out people using chat to ask questions. There is another public R chat room for general chat. Whether we need to follow this model for Cross Validated I'm not sure, but using chat to resolve issues and disagreement in a great suggestion and well made! – Gavin Simpson Sep 18 '13 at 16:34

This has clearly caused a lot of confusion; it seems hard to be specific. However, I think perhaps this is a starting point:

If answering the question requires knowledge of statistics, ask it here. If it does not, ask it on SO.

I think the division should be in terms of who is going to answer. I hang out here (and on English Language Learner and sometimes History) but on SO I almost exclusively ask questions rather than answer them. I am not a programmer. Some questions that are currently getting migrated to SO are ones I could answer - they involve SAS or R but really require someone to know statistics, not programming.

• In principle I agree with this but am not sure this helps the situation in terms of "defining clearer house rules on software related questions". As I said in another comment, I think that some form of this reasoning is used by most (probably all) close reviewers. The problem is that this reasoning obviously leads to different conclusions for different people since it is inherently subjective. – Macro Sep 18 '13 at 14:02
• @Macro I don't think that we will ever come up with a really objective way to do this, but I think my wording may be clearer in the mind of the people asking the questions. Then we can get more questions being asked in the right place to start with. I also think it may make it easier for people who want to answer questions to know where to go. Not perfect, of course, but maybe better than now. – Peter Flom Sep 18 '13 at 14:08
• I see. That's a different way of looking at it. I agree that if higher quality questions were asked to start with, any arguments over closure reasons, etc., would be unnecessary and this problem would sort itself out. In light of your comment - and I have no idea whether or not this is practicable - maybe the site's functionality should be changed so that new users are required (and not just offered a badge, etc.) to read the "How to ask" section before asking. At this point, I'm pretty sure a vast majority do not :-) – Macro Sep 18 '13 at 14:12
• The issue with such simple wording is i) it precludes the fact that the asker may not appreciate the statistical issues, and ii) it grossly over-simplifies what would be acceptable on Stack Overflow. Part of my aims with this Q&A is to put in place better guidelines for users that leave comments about Qs being OT & should be on Stack Overflow, who vote to close in the same manner, who flag questions etc. – Gavin Simpson Sep 18 '13 at 16:39
• @Macro Good point; having better, clearer, more specific(?) guidelines in the How to ask page helps even if people don't read it. Askers of poor questions should be pointed to that resource and given time to respond before being totally dumped on. – Gavin Simpson Sep 18 '13 at 16:40
• @Gavin, your main point that I find persuasive is about "giving them time to respond". Just a couple comments: i) As I said in the other thread, there was a time where I would press the OP for relevant details for a low quality question. After being undermined by other high reps users who "took a stab at it" as is (often admitting in their answer that they're unsure what the question is asking), this gets very discouraging so, for the most part, I stopped. I'd be surprised if I'm the first or last to feel this way. – Macro Sep 18 '13 at 17:39
• (cont.) ii) In my experience, you only get a response from the OP, say, 30% of the time. Perhaps they learn from the other users who will answer "bad" questions and figure they can just wait rather than putting some effort in and iii) With the review queue the way it is, and the relevant badges which can be irresistible for some to "game", you have to make a decision. If you "skip" it, it's gone forever and you get no credit for the review. This dynamic will be hard to overcome (there will always be some users, usually newer ones, who are "gamers"), no matter what policy you try to install. – Macro Sep 18 '13 at 17:42
• @Macro: Indeed, I can assure you, you are neither the first nor the last. :-) – cardinal Sep 18 '13 at 18:56