8
$\begingroup$

I've noticed that, lately, we have gotten a few essentially one-line answers of the form "Look at <insert link>, where good details are provided."

I am wondering what, if anything, should be done to address such answers. For example, do we

  1. Let the voting mechanism handle them,
  2. Address them in the FAQ,
  3. Actively manage and discourage such answers, or
  4. Do nothing (because I'm completely off-base here)?

My reason for asking is as follows.

While, certainly there are many outstanding internet resources that provide better information than could be detailed in a short stats.SE post, it seems to me that the ephemeral nature of much of the web means that a lot of these answers (which may be highly upvoted and/or accepted) may turn out to be essentially useless in the future if the link gets broken. If the goal of stats.SE is to provide a centralized community for addressing statistical problems, then it seems incumbent on us to ensure that the content is as stable as possible.

My inclination would be to actively discourage such answers and suggest that an appropriate synopsis be placed in the body of the post with a link to the (potentially unstable) source. Likewise, for links to peer-reviewed references, answerers should be strongly encouraged to use stable links, as can be found on sites like projecteuclid.org and other online repositories.

What do other community members think?

| |
$\endgroup$
  • $\begingroup$ Would you be able to supply a few representative links? I would like to see what I have been overlooking. $\endgroup$ – whuber Aug 2 '11 at 17:33
  • $\begingroup$ I'll collect a few and post them. I was hesitant to make it seem like I was calling out particular users/answers. $\endgroup$ – cardinal Aug 2 '11 at 17:39
  • $\begingroup$ Here's one example I could find quickly: stats.stackexchange.com/questions/13747/… $\endgroup$ – cardinal Aug 2 '11 at 17:43
  • $\begingroup$ Interesting, because that one is not picked up by the "low quality answers" tool. It therefore is a nice example of the value of community involvement in encouraging better answers: it's possible none of the mods would ever have seen it, or quite some time might pass before we do. $\endgroup$ – whuber Aug 2 '11 at 17:45
  • $\begingroup$ The user has pretty high rep. Maybe the low-quality answers tool filters users out once they get enough positive feedback from the community? $\endgroup$ – cardinal Aug 2 '11 at 17:47
  • $\begingroup$ There are several ways to earn high rep. One of them is to answer lots of questions and have a colleague who upvotes all (or at least most) of your replies. Guess what? Soon you have a few thousand points even if the rest of the community has ignored you. A much better gauge of rep, therefore, is to look at average points per reply. (You can refine this by counting each question as half of a reply, etc., but you get the idea.) Counting high-rep replies is another good method, too. Five or more net upvotes on CV is a pretty good threshold to use. Finally, look at % of accepted answers. $\endgroup$ – whuber Aug 2 '11 at 17:52
  • $\begingroup$ I believe the tool looks solely at the text (of questions and replies) without regard to who wrote them. In the last month low-quality posts by users with ~2K rep have been flagged, too. (On another site I have seen posts by the moderators flagged this way...) $\endgroup$ – whuber Aug 2 '11 at 18:16
  • $\begingroup$ I have always been a bit dissatisfied with the reputation scheme on this site, but that does not mean I have a "better" solution. I realize it serves as motivation for some and gives some very crude measure of "trustworthiness" in content. The information content of rep on stats.SE is generally somewhat higher than on other "sister" sites, in my estimation. My main point in mentioning was to propose a possible explanation for why the aforementioned post didn't show up. But, your empirical evidence strongly suggests otherwise. $\endgroup$ – cardinal Aug 2 '11 at 20:10
  • $\begingroup$ Here is another that may fit the bill: stats.stackexchange.com/questions/13690/… $\endgroup$ – cardinal Aug 2 '11 at 20:10
  • $\begingroup$ Borderline: stats.stackexchange.com/questions/13686/… and stats.stackexchange.com/questions/13686/… $\endgroup$ – cardinal Aug 2 '11 at 20:16
  • $\begingroup$ The first is a good example (#13690). It slipped through the screening tools and I agree it should be amplified. The borderline ones are, indeed, borderline. I saw them both and left them alone because their chief merit lies in identifying a specific software tool and both are kind enough to provide direct links. Often, that's all one really needs in the way of a reply. I feel that some description and explanation would usually be helpful in such cases, but on the other hand I don't want to raise any barriers that might discourage such answers. $\endgroup$ – whuber Aug 2 '11 at 21:27
  • $\begingroup$ Agreed. (A brief discussion of what the elastic net is and why it's relevant to the question could go in one of those.) $\endgroup$ – cardinal Aug 2 '11 at 22:21
  • $\begingroup$ Are Wikipedia links OK? I found myself using them in almost every answer I post, and they probably comprise 80% of the links I provide. In many cases, the original papers are just too cumbersome, and better explanations (implementations, sometimes even notation) has been developed elsewhere (and, with high likelihood, summarized on Wikipedia). $\endgroup$ – StasK Aug 11 '11 at 7:47
  • $\begingroup$ @StasK, my personal opinion is that wiki links can be ok, but it's still good to provide a synopsis of the main point you're trying to make within the body of the post. Also, care should still be taken; I've found some grossly incorrect stuff (without even searching hard!) on some wiki pages on statistical topics. Even worse, one such example ended up referenced here on stats.SE in an answer! $\endgroup$ – cardinal Aug 14 '11 at 14:04
  • $\begingroup$ @cardinal: Oh my. I usually check if Wikipedia makes sense (and edit it otherwise :)). $\endgroup$ – StasK Aug 16 '11 at 0:44
9
$\begingroup$

This is technically covered in

https://stats.stackexchange.com/questions/how-to-answer

Provide context for links

A link to a potential solution is always welcome, but please add context around the link so your fellow users will have some idea what it is and why it’s there . Always quote the most relevant part of an important link, in case the target site is unreachable or goes permanently offline.

I generally try to edit in context when I see answers that are little more than a link with no context.

| |
$\endgroup$
  • 7
    $\begingroup$ I usually ask the respondent to do those edits, because they are the experts and they know what to emphasize in their descriptions of the links. $\endgroup$ – whuber Aug 3 '11 at 13:51
6
$\begingroup$

I was guilty of this behavior in one of my first replies :-).

As a mod, I hunt for such low-quality answers routinely (there's a tool that flags them temporarily) and I then post comments encouraging respondents to summarize or amplify. There's no good mechanism for me to track these, though, so I don't always follow up in a timely manner. (It's only polite to wait at least a day because not everyone visits the site all day long.) When I do come across them again, if no change has occurred I usually downvote or vote to close, depending on how uninformative the reply has been.

I would love to see more high-rep community members step up and take similar actions where warranted. Very few members flag or vote to close posts any more--or even downvote them, which is an excellent tool for encouraging changes in behavior. (I figure that reflects faith in your moderators :-), but still...) Just please remember to use such tools in a kind fashion by accompanying their use with positive, constructive comments.

| |
$\endgroup$
  • $\begingroup$ I've noticed that you and the other mods have been quite proactive about this the last couple months or so. $\endgroup$ – cardinal Aug 2 '11 at 17:40
  • $\begingroup$ Hm which tool flags the low quality answers? Is it available for non-mods? My reading behaviour is highly skewed towards new questions, where I either edit the question, suggest more clarification, or answer it myself. If there is a feeling that help is needed for policing, I will be glad to change my usage patterns. As for flags and votes to close, somehow I did not see anything worthy of a flag recently. $\endgroup$ – mpiktas Aug 3 '11 at 10:44
  • 1
    $\begingroup$ @mpiktas: Look under the review link at the top of the page to the right of your user name when you are logged in. $\endgroup$ – cardinal Aug 3 '11 at 12:10
  • $\begingroup$ @cardinal, this is for the first time posters. I thought whuber was talking about general tool not only for first time posters. $\endgroup$ – mpiktas Aug 3 '11 at 12:23
  • $\begingroup$ @mpiktas: There are four subtabs there to choose from. The second is low-quality posts. Maybe the mods toolkit is different, but I suspect they are at least related. :) $\endgroup$ – cardinal Aug 3 '11 at 12:27
  • $\begingroup$ @cardinal, whoa, I did not notice the subtabs! Will be using this a lot from now on :) $\endgroup$ – mpiktas Aug 3 '11 at 12:29

You must log in to answer this question.

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