As I assume that many of you are aware, there are no [soft-question], [advice], or [career-development] tags in stats.SE. I was wondering why this was, and who decides it?

Some of the most successful / popular questions of math.SE have the [soft-question], [advice], or [career-development] tags which can be see from just looking at the first couple of pages of math.SE.

I can see no reason why these types of questions should be important to people on math.SE but not on stats.SE. Moreover the argument is that these are "not encouraged by the format of SE" seems strange as they feature prominently on one of the most successful SE sites. (One could generalise this argument to math.overflow, although that is not part of the SE network.)

  • 6
    $\begingroup$ They don't exist because nobody created them. That doesn't imply any kind of prejudice; it's simply that nobody has felt the need to do so. (Given that nobody has felt the need in over three and a half years and almost 33 thousand questions, that's a pretty good argument that they aren't needed.) Personally, I feel the 'soft question' tag is a bad idea for several reasons (for example (i) it introduces an utterly subjective category that (ii) indirectly labels some questions to be less important/valuable than others). Statistics is not simply mathematics and needn't use the same sets of tags $\endgroup$ – Glen_b -Reinstate Monica Mar 7 '14 at 1:07
  • 4
    $\begingroup$ Not all rules are equally enforced on all sites, & I am not part of the group actively participating in math.SE. However, [soft-question] is clearly a meta tag & SE policy is that meta tags are explicitly discouraged. [career-development] seems fine, if people wanted to ask questions about that, [advice] seems ambiguous, but [soft-question] I would be against. $\endgroup$ – gung - Reinstate Monica Mar 7 '14 at 2:20
  • $\begingroup$ why the downvote? I am asking a question, surely disagreeing with the question is not a reason for a downvote? $\endgroup$ – hmmmm Mar 8 '14 at 11:45
  • 6
    $\begingroup$ Don't take it personally, @hmmmm. I am one of the downvoters. I rarely downvote on CV, but do much more frequently on meta.CV. Downvotes have different meanings on the 2 sites: on main sites, it means you think the answer is wrong or poorly done, etc; on meta sites, it means you don't agree w/ the suggestion, but it doesn't imply that it was poorly done or incorrect or some such. $\endgroup$ – gung - Reinstate Monica Mar 8 '14 at 16:16
  • 1
    $\begingroup$ See also meta.stackexchange.com/q/224759/246931 about the soft-question tag on MSO. $\endgroup$ – Nick Stauner Mar 10 '14 at 1:59
  • 2
    $\begingroup$ Downvotes on meta don't mean it was a bad question, they generally mean "I don't think this idea is a good way for the site to go". They're certainly not personal. (I didn't downvote - but I didn't upvote what is a perfectly reasonable thing to ask about, for the reason that it would tend here to indicate agreement) $\endgroup$ – Glen_b -Reinstate Monica Mar 11 '14 at 0:21

Tags are created by people who ask questions. To date--after three and a half years and some 35,000 questions--nobody has seen fit to apply such a tag. None of the active high-rep community members have seen any need to create and apply such tags retroactively. Although I would not necessarily object to their creation, I see little need for them within the network of tags that has arisen within this site. I do have some concern that tags with "soft" in them could be viewed prejudicially, as if such questions somehow had less merit or interest than "hard" tags related to techniques and theory.


Just as other folks have said, statistics is not math, and there are sufficiently many differences in how the two areas project to academia, the real life, and the SE system. My impression (based on limited excursions to math.SE) is that most questions on math.SE come from people who are studying math for their own enjoyment. The distribution of the question sophistication is bimodal: 90% of questions are graduate-level questions, and the remaining 10% of the questions are high school calculus.

Stats.SE sees visitors from computer science, economics, psychology, among others, so plenty of questions on stats.SE come from people who were thrown into statistics and had to produce a specific analysis for their report, dissertation or paper, with no intention of making statistics their lifetime passion and a career choice. They could be hit-and-run user####### folks who bring the discussion forum expectations into their formatting of the questions and answers, use of tags, etc. The level of sophistication of these questions is essentially "What does it mean to compute $\int_0^\infty x^2 {\rm e}^{-x} \, {\rm d}x$ by parts?", and there's a fair amount of noise because the different disciplines that use statistics (or machine learning, or feature extraction, or data science, or quantitatve methodology, or whatever the equivalents of statistics are in the disciplines) use their own idiosyncratic names for the procedures that they need to produce. This is all to say that there is little room to soften this site any further, and that's why the career-related tags see little demand. The distribution of sophistication is smoother on stats.SE, I think, but this may be an artifact of the fact that you need to have a college degree to begin with before anything statistical would come your way, so you would have a need to address stats.SE.

Finally, the communities may have developed some conventions of what they mean by advice or soft-question. We have some high popularity advice-type questions about statistical tools (Python?; SAS vs. R?; skillset of a statistician; etc.), which don't ask specific technical questions, but instead ask broader why or what-to-do types of questions. A smart reader thinking about building their career in statistics would gain a lot from reading these. I guess we just did not have in our conventions that such questions should be tagged advice.

  • 2
    $\begingroup$ See meta.stats.stackexchange.com/users/5739/stask for a frank (but in my view fair) comment about the use of user# handles. $\endgroup$ – Nick Cox Mar 12 '14 at 9:30
  • 2
    $\begingroup$ I sense more questions from people taking introductory statistics courses at undergraduate level than is implied here, but if that's true it strengthens the argument here, as such people are not thinking about jobs or careers, just the last lecture or the next assignment. $\endgroup$ – Nick Cox Mar 12 '14 at 10:16
  • $\begingroup$ I was wondering what you meant by "studying math for their own enjoyment", do you mean whilst undertaking an Msc/Phd or some sort of independent home study? $\endgroup$ – hmmmm May 4 '15 at 14:18

First, we do have (and also , , , and ). It's not a particularly popular tag here, as you can see. It may not always be applied where it should (e.g., Things to consider about masters programs in statistics), but we're not especially zealous about retagging old questions here (see Retagging old questions for more nuanced discussion).

I agree with @gung that an tag sounds overly ambiguous. I'm not sure what it would add to organize questions this way, and also wonder whether it's really doing math.SE any major favors. Feel free to pitch it to us if you think it has worked over there, but popularity isn't necessarily an indication of practical utility. It seems particularly fishy to me that its tag wiki suggests using another tag too, but hasn't been filled out much at all yet – not even to disambiguate it from the other, the tag.

As for that tag, I agree it could be misunderstood or misapplied as pejorative derogation of a question's scientific merit, but tags are misapplied often enough (gung had to correct me on my own use!) that I prefer to accept this as par for the course and worry more about the intended meaning of the tag. There's still plenty to worry about there. First, you've got another excerpt-only tag wiki there, so its use doesn't seem very clear to begin with – not exactly a strong candidate to nominate over here...Second, "questions that don't admit a definitive answer" reminds me a lot of the Help Center's don't-ask page, as well as the too broad text – "There are either too many possible answers..." – and the primarily opinion-based text for closing questions:

Many good questions generate some degree of opinion based on expert experience, but answers to this question will tend to be almost entirely based on opinions, rather than facts, references, or specific expertise.

I'm concerned that a tag would legitimize questions that test these boundaries. Maybe this wouldn't be a bad thing; the questions could help define the boundaries, and the tag could help mark boundary cases, but this begins to approach the dangerous, implicative, prejudicial territory again.

Third, I echo @Glen_b's observation that it's quite a subjective call as to when one should expect a definitive answer. Since statistics often entails a considerable amount of error tolerance and choice of competing models, definitive answers may be somewhat rarer here than on Mathematics. I would prefer not to see questions organized by this characteristic, but by its opposite (, I assume) wherever a definitive answer is clearly necessary. IMO, if the idea of a tag is to encourage less-than-definitive answers, the place to do that is in the body of the question, not in the list of tags.

BTW, I downvoted too for these reasons, but to echo gung once more, it's not because I found your question useless or unclear! As a matter of fact, I just retagged a new question with thanks to having that tag brought to my attention...so thanks for asking anyway!


You must log in to answer this question.

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