Recently I've had some discussions with @whuber about what constitutes appropriate use of the 'Not an answer' flag where I was referred to this thread:

To flag or not to flag and a request for auditing

which essentially says that the only appropriate use of the 'Not an answer' flag are answers like "Thanks!" or "I have the same problem". However, in another thread on a related subject a moderator on meta SO weighs in:

What should I *not* use flags for?

where it is indicated that is it, in fact, appropriate for moderators to act on flags raised due to a completely irrelevant answer. Let's look and what is written in the flag box next to "Not an answer":

"This was posted as an answer, but it does not answer the question. It should possibly be an edit, a comment, another question, or deleted altogether."

How we go from that to 'only "Thanks!" and "I have the same problem" responses fall into that category' reflects a bit of a disconnect to me.

The discussion began on this thread: How much does quantizing continuous data affect power? where the question is asking about the consequences of categorizing continuous data according to the quantiles. The only answer (which I flagged and was declined) was about quantile regression, which relates to statistics but has nothing to do with the question. Another example that sparked my interest is here.

I'd like to open up a discussion about what exactly is wrong with something like this example being deemed "not an answer" and how that can be reconciled with the text written in the box under the 'not an answer' flag, possibly using the question above as an illustrative example. Perhaps the text should be changed? Or, the general consensus on what is and is not an answer needs to be clarified?

Edit (5/16/12): After much discussion it appears the disconnect is that I'm approaching this with a very literal interpretation of the rules cited whereas the moderators approach this problem with more of a "spirit of the rule" type of interpretation, which leads to a cases that may be "borderline" in the sense that a rational individual may think an answer does fit with the spirit of the rule while others do not. This approach is less clear cut but may be for the best.

| |
  • 1
    $\begingroup$ Please note that the referenced thread is part of the SE meta FAQ: it reflects sitewide policy, not just CV policy. I cannot speak for the other mods here, but I tend to mark any flag as "helpful" unless it absolutely has to be declined, because I want to encourage people to flag things when appropriate and there are a lot of gray areas. So please don't take these "helpful" responses as reflecting any disagreement or confusion among mods. $\endgroup$ – whuber Apr 3 '12 at 17:49
  • $\begingroup$ OK. But, in your opinion, apparently, the flags I raised simply had to be declined, whereas similar ones in the past have been deemed helpful (and often times resulted in deletion of the answer) - this seems like disagreement among the mods. What am I missing? $\endgroup$ – Macro Apr 3 '12 at 17:59
  • $\begingroup$ Perhaps only an evolution in my understanding of what is an appropriate action to take :-). Mods undergo (voluntary) regular training to help us understand policies, deal with common issues, etc. One thing that has become clearer to me over time--as explained in the referenced threads--is that flags are for problems. Use your votes to indicate good or bad answers. Overall, this community tends not to vote, which is a real problem. We get about 1/2 the upvotes per question per user seen on other sites and almost nobody downvotes bad entries. Use your votes, early and often! $\endgroup$ – whuber Apr 3 '12 at 18:03
  • $\begingroup$ I understand, @whuber. In that case, I'm just trying to reconcile the disconnect between what is written in the flag box under "not an answer" and what is written in the SE meta thread I linked to. Obviously, I think an answer that has nothing to do with the question is... not an answer... It's not an answer which represents some misunderstanding of the subject at hand, which is what I thought downvotes represented. It's like the difference between recommending the wrong test for the right problem (downvote) or the right test for an unrelated problem (not an answer) - this was my understanding $\endgroup$ – Macro Apr 3 '12 at 18:13
  • 2
    $\begingroup$ @Macro "I'm just trying to reconcile the disconnect between what is written ..." and "this seems like disagreement among the mods ..." are really two different things. I share whuber's point of view: I mark flags as "helpful" when they suggest inappropriate responses ("thanks!", "I have the same problem...") that I delete immediately, but also when they catch my attention onto a potential issue with someone's reply; in the latter case, I won't necessarily act by deleting the related post. (...) $\endgroup$ – chl Apr 3 '12 at 21:01
  • $\begingroup$ (Con't) You can vote down and leave a comment to indicate what's wrong with the response. Other users can vote on your comment and the response as well. Few responses with -5 votes have survived here. -- Edit: Ah, and now that I've read whuber's response below, I think he developed much better than I did the above point (voting vs. flagging). $\endgroup$ – chl Apr 3 '12 at 21:03
  • 1
    $\begingroup$ OK, chl. "this seems like disagreement among the mods ..." was not meant as a focal point and seems to be taking away from the real question so I've deleted it. $\endgroup$ – Macro Apr 10 '12 at 12:27
  • 4
    $\begingroup$ @whuber, the issue with upvotes (and even more so with downvotes) per user is that there may be really few people who (think they) are in a position to judge on the quality of an answer. People come to CV to ask a question they have been desperately trying to answer offline using the networks of colleagues, advisers, etc., and at any given time, at least half of the questions on the first screen are from users with rep 1. They are unlikely to have the culture that a Yearling or a Marshal would have. Other SE sites may have a relatively higher proportion of experts and enthusiasts among users. $\endgroup$ – StasK Apr 13 '12 at 16:19
  • 3
    $\begingroup$ @Stas Let me share some data with you :-). We average 150-200 upvotes per day. We have about 170 1K+ users, 373 "yearlings," and 151 "enthusiasts." This suggests between 150 and 300 people revisit the site several times weekly, if not daily. Thus, each avid, regular user casts an average of 0.5 to 1 vote per day. In fact, the distribution is highly skewed, as the "Users" tab shows: only about 30 people--around 10-20% of the core community, however you look at it--cast almost all these votes. There's room for improvement. $\endgroup$ – whuber Apr 13 '12 at 16:31
  • 2
    $\begingroup$ @whuber, what you've just said seems like evidence that the sloppy things I've been flagging (that seem, by an reasonable interpretation, to fall under the written description of the 'Not an answer' flag) won't get taken care of "on their own". $\endgroup$ – Macro Apr 13 '12 at 16:45
  • $\begingroup$ Good point, @Macro. The natural follow-up to the implied question about who should be tending the "sloppy things" (that is, incorrect or ambiguous or misleading replies) is "if not the community, then who?" The answer is not the mods and not the SE team. (This is just how it is: I'm not stating a personal opinion.) On SE, it always comes back to the community: information about the quality of Q's and A's has to come from our collective votes and comments. That's a great argument for why community members should vote far more often than they actually do! $\endgroup$ – whuber Apr 13 '12 at 16:52
  • 2
    $\begingroup$ OK then is it possible to resolve why the 'Not an Answer' flag indicates, verbatim - "This was posted as an answer, but it does not answer the question. It should possibly be an edit, a comment, another question, or deleted altogether." when this is not how it's enforced? It seems like it's someone's responsibility to enforce that and no one has the power to except the mods. Again, I'm talking about when the answer, while not being "Thanks!" or "Me too!", doesn't remotely answer the question. To me, it's clear that this falls under the verbatim definition above. Where am I going wrong? $\endgroup$ – Macro Apr 13 '12 at 17:01
  • 2
    $\begingroup$ I have reviewed your flag history, Macro. What it shows is that enforcement of the "Not an answer" flag has been consistently as I describe it: only the replies that are truly non-answers have been deleted. Those that were merely wrong or off the mark were not deleted. What may be confusing you is that in half a dozen cases, you flagged a reply as "not an answer"; those replies were not deleted (and still stand), indicating the flag really wasn't deemed correct; but your flag was marked "helpful" by a moderator. This is likely because we appreciate the help and want to encourage flagging. $\endgroup$ – whuber Apr 15 '12 at 15:10
  • 2
    $\begingroup$ @Macro, I won't be available to chat in real time most of this week, but we (and anyone else who cares to join) can still continue this conversation in the chat room at chat.stackexchange.com/rooms/3128/what-is-not-an-answer. Liberal use of the "@" construct will help notify people when new material is available and awaiting their reply. $\endgroup$ – whuber Apr 16 '12 at 21:25
  • 2
    $\begingroup$ Perhaps it's obvious, but reformulating 'this is not an answer' into 'this is not the kind of response that could be an answer to this or any other question' makes things perfectly clear. Also, I actually never noticed I could do this, so I shall now happily become part of the problem... $\endgroup$ – conjugateprior Apr 17 '12 at 13:01

Flagging and voting

The comments after the question ultimately focus on what does "not an answer" mean and what is the purpose of flags. Perhaps this can be clarified by pointing out that moderators do not adjudicate the correctness of answers. We would like to have that breadth of knowledge, but I'm sure I don't and I suspect the other mods aren't omniscient, either.

So what the site is asking you, the community, to do, is to signal correctness, appropriateness, usefulness, suitability, and all those good things, by voting. If you understand a reply (or a question), whether it's good or bad, then please vote on it! Don't leave any of our pages without at least considering placing a vote on the question and at least one reply (if not all of them). As a statistically minded person, you know that a small number of votes is a highly uncertain indicator, whereas a large number--although imperfect, to be sure--tends on the whole to be more reliable. Your vote counts. It is the means by which reputation is created, more power is given to the community to act, and the quality of the site is maintained.

Moderators are here to help when there is a functional problem with some element of the site. These are usually posts that either contain no information ("thanks, I liked that"), are misplaced (such as questions posted as replied), obnoxious (spam, flames, etc.), or simply incomprehensible (such as problems with the language, corrupted text, possible site bugs, etc.). When you find such a problem, please flag it.

Yes, sometimes it's hard to distinguish a functional problem from a bad question or reply: if a reply is incomprehensible, should it be downvoted or deleted? (My tendency is to delete it whenever it becomes apparent that a large number of downvotes will be accumulated, out of compassion for the responsible party and their reputation. I could easily be persuaded to adjust that policy: this is one of those gray areas.)

I would suggest that if you think a reply is wrong or misdirected, then do some combination of (1) providing a constructive comment, (2) downvoting, (3) providing a counter-reply, (4) inviting the poster to discuss the issue with you in a chat room, and (5) leaving a comment encouraging other users to vote appropriately (in that order, as needed). When this is done in a civil manner, moderators do not need to be involved and in fact won't be of any special help (except perhaps to help create a chat room or perform other supporting functions like that).

Consistency of moderation

The modus operandi of moderators are not completely uniform: we work in slightly different ways. For example, some of us prefer to edit posts more than others; some of us post more comments (to encourage people to do their own editing, closing, deleting, etc.). We have different levels of experience with other sites in the SE universe, which can affect our tendencies to migrate questions. However, in the course of our work we naturally see much of what each other does and we check for inconsistencies. We confer about problematic situations with each other (in a dedicated chat room) or with more than 200 mods across all the SE sites (in another dedicated chat room). We bring these conversations public here on meta (or occasionally in comments) whenever we think the community might appreciate getting involved.

This constant contact helps assure a consistency in moderation at any given time, but it also leads to our evolution as a group as we learn from previous situations, mods on other sites, and the SE team. Therefore you will indeed find inconsistencies between actions taken a while ago and today--but I hope that today's actions tend to be better and more consonant with SE policies overall than yesterday's were.

| |
  • 1
    $\begingroup$ Re: "moderators do not adjudicate the correctness of answers." I was never asking you to make a judgment about the correctness of an answer. Once again, I'm only interested in cases where the answer, by no stretch of the imagination, answers the question. For example, if someone asks about discretizing data by using a quantile split, and I respond with an answer about quantile regression - this may be a fine answer... to a completely different question that has nothing to do with the question at hand. $\endgroup$ – Macro Apr 10 '12 at 12:24
  • 1
    $\begingroup$ I still don't feel the conflict with "This was posted as an answer, but it does not answer the question. It should possibly be an edit, a comment, another question, or deleted altogether" and what you've described as the way flagging/moderating happens in practice has been resolved (or even addressed, for that matter). $\endgroup$ – Macro Apr 10 '12 at 12:25

Let me share some thoughts on this thread.

I recently dismissed one of your flag on a reply to Standard deviation of standard deviation. You may have noticed that I added a personal message explaining why. Below is a copy:

enter image description here

The incriminated post can be considered as a response as it specifically points to a reference that might not be easily found with naive Google queries, at least for someone totally new to the subject. Users coming across this thread may well find in this reply a first start. Some of the other responses on this thread, including yours, demonstrate that more technical points can be made. Now, whether this is a good response or not is a matter of voting, not flagging.

About "barely more than a link".
List of possible reasons for post removal is headed with

Answers that do not fundamentally answer the question may be removed

(emphasis is mine)

This site aims to build a collection of self-contained and sustainable responses to specific questions. "Dangling" links or pointers to ephemeral resources should be avoided; link to authoritative references are to be preferred (I'm not saying that Wikipedia falls under this category). If a reply correctly points to a perennial source (that's the case for Wikipedia), and answers the question in an unambiguous manner, then I do not see why it would not be judged as a valid response. If users do not consider it as a good response, it is likely it won't attract many upvotes. (We even have badges for highly contrasted votes on question and answer.)

I for one prefer more elaborated replies, even if they are built around an external reference (when it is appropriately quoted). But I am not the only user on this site, and other users may express a different opinion on what constitutes a good response, when to vote up, etc. Let's go for it: community voting is not perfect, but it certainly helps to filter good replies. If you feel that an answer has low quality, you can leave a comment and/or propose a better response, and express your opinion with your vote. I hope this point was made clear by @whuber in his first reply.

In this particular case, the supplied answer (from a very active user in the first 6 months of CV life) seems acceptable, yet it might be improved. Users with 2k rep can edit posts, by the way. Community edits are one way to enhance the quality of this site. In all cases, community voting is the way to regulate a thread or moderate older ones to account for new replies.

| |
  • 1
    $\begingroup$ Thank you, chl. This is a concise, respectful description of your disagreement with my flag there. I think this is a clear description of when "nothing more than a link" can be a serviceable answer. I happen to disagree that the flag you just brought up was a serviceable answer, but I accept the 'declined' as reasonable. My position on this is evolving from "strict adherence to what is posted verbatim in the guidelines" to something less well defined. $\endgroup$ – Macro May 16 '12 at 15:07
  • $\begingroup$ FYI: The way I'd been handling these kinds of flags was pretty closely in line with whuber's answer here: meta.stats.stackexchange.com/questions/912/… ... Based on the number of such flags of mine that have been declined, I suppose the mod's view on this subject has evolved as well. In any case, as I said, I'm modifying my threshold of what is "flaggable" $\endgroup$ – Macro May 16 '12 at 15:18
  • 4
    $\begingroup$ Macro, because we have mutually agreed to suspend the long comment thread accompanying my second reply, and because you continue to object to the reply itself and I don't want to offend, I have deleted that reply. I am grateful to @chl for weighing in here and am glad to see your constructive reactions to his points. $\endgroup$ – whuber May 16 '12 at 16:02
  • 5
    $\begingroup$ very good, @whuber. I'm realizing now that all along we seemed to be arguing completely different points. While I harped on a literal interpretation of the FAQ, you argued for more of a "spirit of the rule" type of interpretation, which my view on is evolving. My apologies if my flagging bugged you but my intentions were not malicious - I'll trust your intentions in the comments I objected to also were not. Now, let's get back to using this site to provide high quality statistical advice... $\endgroup$ – Macro May 16 '12 at 16:15

You must log in to answer this question.

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