I find that many questions go unanswered for lack of a great answer. That applies to both CV and here on CV meta. Whether this is due to brain drain, or a flood new users, or perfectionism, is hard to say.

Moreover, a lot of ill-posed questions amount to the asker not knowing enough about the field to set them in the right direction, and in the review queue there are plenty of questions that are clear enough to answer, but not clear enough to answer well. Edit: Other times, the question has an answer, but it seems that none of the expert users on the site know the answer.

Personally, I know a little about a lot, and a lot about a little. If a question doesn't attract a good answer, should I answer it with my own sorta-okay answer? (Example: https://stats.stackexchange.com/a/364232/36229) I do my best to be as not-wrong as possible, but is being not-wrong good enough, or should we only answer questions where we can really knock it out of the park, and ignore or close the rest? My opinion is that the former is preferable, but that might be at odds with how others' feel.

Rather than starting on some kind of vigilante rampage through the Close Votes queue -- is there an official position and/or community consensus on "okay, but not that great" answers?

• Closely related: stats.meta.stackexchange.com/questions/1282. It might also behoove us to review stats.stackexchange.com/help/how-to-answer from time to time, especially the section beginning "Not all questions can or should be answered here." – whuber Aug 27 '18 at 20:16
• Could you expand on, or give examples of, the kind of difference you see between "quick-and-dirty"/"sorta-okay" & "great"/"good"? I think it's hard to answer you without a fairly clear notion of that - "Whatever is worth doing at all, is worth doing well" vs "The best is the enemy of the good". – Scortchi - Reinstate Monica Aug 27 '18 at 22:00
• @Scortchi here's a "sorta okay" answer of mine from today: stats.stackexchange.com/a/364232/36229 . Someone with even moderate practical experience in CNNs would be in a position to provide a better answer. – shadowtalker Aug 27 '18 at 22:06
• In my previous thread about how to improve the answer rate, I advocate writing brief answers to questions which can be completely, and clearly answered briefly. This sounds a little tautological, but it’s motivated as a reaction against writing answers in comments. Personally, I write an answer only when I’m confident that it addresses the post. That threshold will be different for everyone, but I think it’s a reasonable standard. – Sycorax Aug 27 '18 at 23:04
• I tend to comment quite extensively at times and potentially interact with the OP. If that seems to help the OP I might go ahead and take the time to make it into a stand-alone answer. I find it faster and easier. – usεr11852 Aug 27 '18 at 23:12
• Forgot to include the link -- stats.meta.stackexchange.com/questions/5325/… – Sycorax Aug 28 '18 at 17:56
• @whuber This thread reminds me of your words back from 2012: "So when deciding whether to spend your valuable time rushing to answer a lot of questions or answering a single one really well, I urge you to consider expressing your expertise and insight in a high-quality, well-crafted answer whenever you can." (stats.meta.stackexchange.com/a/1248). Personally, I tend to agree with you, but I see that many people (e.g. in this very thread) worry about our low answer rate and seem rather to advocate an opposite approach. – amoeba Aug 29 '18 at 19:57
• @amoeba You have me pegged correctly. – whuber Aug 29 '18 at 22:05
• @amoeba I don't see this as an either/or situation. I think the optimal community strategy is a mix of experts who write fewer long, well-crafted answers and journeyman statisticians (to use Sycorax's phrase) who write many short answers. The short answers can take some pressure off the experts, and let them focus on writing detailed answers. The only alternative is to recruit a huge number of new experts, which seems unlikely to happen soon. – mkt - Reinstate Monica Aug 30 '18 at 7:53
• (Well, we could also choose to leave a lot of questions open but unanswered, or to close many more by being a lot tougher. These options seem to have less support.) – mkt - Reinstate Monica Aug 30 '18 at 8:04

I come down strongly in favour of providing quick-and-dirty answers, given two conditions: i) the question is not a bad one (in which case, close and/or downvote), and ii) the alternative is giving no feedback of any kind (comments asking for clarification are therefore an exception). As for why Q&D answers are worth providing:

1. Someone out there needs help. A partial answer may quickly address their main point or put them on the path to solving it themselves. No answer is no help at all (unless you're giving feedback in comments, of course). This reason doesn't apply if the goal is to develop a shared library of great answers. In practice, I think many of us are also motivated by a general desire to help people with problems. So I'm happy to write a short answer to help out, even if it's not going to be a general reference in the future.
2. The voting mechanism exists to sort out the bad from the okay and the good. If a Q&D answer I provide is unhelpful or misleading, people are welcome to downvote it. And if it gets upvotes, it's a signal - if a noisy one - that it was helpful. This is partly why I think we should be downvoting more in general - it helps send a clearer signal.
3. Providing a quick answer is not a barrier to someone providing a more comprehensive or better one.
4. You mention knowing "a little about a lot, and a lot about a little". In contrast, I know a little about a moderate amount, period. But this little is still enough to be helpful for many simple questions. If I, and other novices like me, can take care of the basic questions, then the true experts can spend their time crafting the great answers that will form the general reference we all desire. It may not be the most efficient use of their time to deal with the basic questions.
5. Related to the previous point, if this type of short answer were more common, it could motivate broader participation - with the end result that we eventually have more people writing the great, detailed answers too. As a newcomer, it was difficult and time-consuming to find a question that was a) well-formulated, b) not a duplicate, and c) within my ability to answer in a detailed manner. I suspect that if the community viewed Q&D answers positively, that would make it easier for new users to start contributing to the community.
• +1, though I would say a quick answer would be useful even if comments are present, if only to get the question out of the "unanswered" list so people working through that list don't have to spend time on answered-in-comments questions. If the comment is an answer, turn it into an answer. If the comment is an answered request for clarification, write an answer. If it is an unanswered request for clarification, close as "unclear". – Stephan Kolassa Aug 29 '18 at 6:44
• @StephanKolassa Agreed. What I meant was a bit different, though poorly phrased: unless I'm personally going to comment to ask for clarification (or flag/vote to close), I feel fine with posting a short answer. – mkt - Reinstate Monica Aug 29 '18 at 10:47
• +1, quick and dirty answer is better than no answer. Not everyone has the time to invest in long encyclopedic latex-supported answers and a complex approach is not always the best approach. – Digio Sep 6 '18 at 11:32
• +1, People are asking for help and providing a quick answer doesn't stop someone else from proving a better answer later. Well said. – SecretAgentMan Sep 6 '18 at 17:16

1. I believe that if a question can be completely answered briefly, then a brief answer is fine.

2. I also believe that the purpose of stats.SE is to provide a long-term repository of statistics$^*$ knowledge.

3. I don't think that there is necessarily any tension between these two points. Sometimes, people ask questions with simple answers:

• the way that you get labels for data is by having a human review the data and label it.
• the way you figure out if $L^2$ regularization improves a neural network is by conducting an experiment;
• the formula for Xavier initialization isn't complicated;
• neural networks depend on the training data;
• "loss function" has many synonyms.
4. None of this changes that questions which are overly broad, unclear, off-topic, or duplicates should be closed as such.

5. I do not believe that short answers are necessarily lower-quality.

One might summarize these points as "A good answer is only as long as necessary," or "Sometimes, short answers are comprehensive answers."

If we only write lengthy answers, there is a gap in answer coverage.

• If the question is on-topic and not a duplicate, it will remain open.

• But it will also remain unanswered if the answer is so straightforward that two sentences suffice and there is a taboo against short answers. Unanswered, open questions serve no useful purpose.

• If there is a perception that the answer box is solely for five-paragraph answers with equations, figures and citations, then questions are instead answered with comments. This isn't what comments are for, nor is it how SE is supposed to work.

• A consequence of not providing appropriately brief answers is that we have a very large & widening gap between questions and answers.

I don't think that short answers are inherently in tension with any of the goals of stats.SE. I just think that because we have so many high-quality answers which are of some length, that has created an impression that the only good answer is a lengthy answer. I'm only advocating for a slight adjustment in how answer-writers go about their task, not a shift in how the platform works.

$^*$ My definition of "statistics" extends to include what many people call "machine learning" and "data science."

I am a firm believer in Pareto's principle, and a quick answer that is 80% correct is better than no answer that is 100% correct. I do not believe that StackExchange aims for Utter Impeccability. I like to say that time doesn't grow on trees, so it's better to give an answer that helps the poster, rather than not give an answer at all.

Anyone who disagrees with your quick-and-dirty answer can downvote it, or post their own answers. Some of our more interesting threads have exactly this kind of competing answers.

EDIT. To clarify, I am not advocating answering bad questions. Bad questions should be closed as "unclear" and the poster encouraged to improve them. (Yes, in this order, to avert answers while the question is yet unimproved. An answer to an ambiguous question will only serve to confuse later readers.)

Rather, I am talking about the kind of question that is clear, though maybe not very interesting. The kind that a student might ask in dealing with their data. These typically do not lend themselves to "great" answers. Frequently, there is nothing overly deep to them. Writing a "great" answer, say by pointing out background theory, or generalizations, will likely go over the head of the asker - who I assume is usually not a statistician, but a consumer of statistics - and rarely help anyone later on.

We get tons of such "bread-and-butter" questions every day. These are completely legitimate, and many of them can be answered quickly by any halfway competent statistician.

The question then is whether to write short answers to five of these, or a long answer to one in the same time. And this is where I come down firmly on the side of writing five short answers that have a chance of helping five people. Maybe one out of these five will ask for clarification, and then I can invest more time in editing and improving that specific thread. Much better than leaving four questions unanswered and writing a detailed answer on one question, which may well be far more than the OP wanted or needed.

This position of mine may be influenced by my working in an agile software development environment. We'd rather build something rudimentary, get feedback and improve incrementally, rather than try to build a large intricate piece of software in one go. (And, to repeat: the requirements to develop against need to be clear before the first unit test is written.)

I understand that people may see this differently and advocate writing fewer, more careful answers. This may be more in keeping with the SE "repository of knowledge" aim, but it will lead to (even) more unanswered questions, which the community seems to believe is a problem. And, as I said, it won't help the psychology grad student who just wants to know how to deal with her data.

• I think @whuber disagrees with this approach (or at least he did 6 years ago): "So when deciding whether to spend your valuable time rushing to answer a lot of questions or answering a single one really well, I urge you to consider expressing your expertise and insight in a high-quality, well-crafted answer whenever you can." (stats.meta.stackexchange.com/a/1248). – amoeba Aug 29 '18 at 19:55
• (-1) I couldn't disagree more with this answer. We are not Twitter or Reddit. We're not here to toss out one-off answers to badly formed, ambiguous questions. Even great answers to ambiguous questions--unless they clearly reformulate the question, which is rare--are a lasting nuisance due to their potential to confuse future readers. The lesser evil is to leave those questions unanswered. – whuber Aug 29 '18 at 22:06
• @whuber: I don't quite see where I advocated answering "badly-formed, ambiguous questions". – Stephan Kolassa Aug 30 '18 at 6:55
• @amoeba: thank you for that link. It looks like whuber and I are on opposite ends of this tradeoff spectrum. (And I am still not advocating tossing off answers to bad questions.) – Stephan Kolassa Aug 30 '18 at 6:59
• I have edited my post. To be honest, I have pontificated often about closing, downvoting and generally manhandling bad questions that I am a bit mystified how at least two people could misunderstand me as advocating answering these. Hope my edit helps. – Stephan Kolassa Aug 30 '18 at 7:36
• Your clarification helps greatly, thank you. Another solution to consider--and in most cases it will lead to superior results--is to perform the research needed to identify existing answers on the site and vote to close the unanswered question. – whuber Aug 30 '18 at 12:28
• @whuber: I agree that we often have close duplicates to "bread and butter" questions. The problem is that while it may be obvious how to adapt a proposed duplicate to a new question for a statistician, this adaptation may be non-obvious to a non-statistician. Plus, it's hard to find a duplicate in over 120,000 questions. – Stephan Kolassa Aug 30 '18 at 14:46
• @Stephan I agree: that's why it's so valuable to close a question as a duplicate. It typically reflects the results of a nontrivial search. – whuber Aug 30 '18 at 16:59