7
$\begingroup$

There is first a question within this question, and, that is "Where should I be posing the question?"

There are certainly people on this site with marvelous minds. The second question, the primary one, relates to the fact that it's nice to have a reputation on Cross Validated, and, many of us, especially those in the "leagues" with 200+ reputations have considerable knowledge of things statistical.

Is there then some measure of reputation, e.g. quarterly ranking, on Cross Validated that is well correlated to some measure of ability to use statistics, e.g., educational level in statistics?

Now the examples (the e.g.'s) I used are somewhat arbitrary, and in the extreme I am asking something along the lines of determining what a "pseudo-PhD" in Cross-Validation might be.

$\endgroup$
48
  • 10
    $\begingroup$ If you take the narrow sense of "reputation" as being total points, it might help to understand that it compresses many variables onto a single scale, including activities to maintain the site, quality and number of questions, giving of bounties, and pure number of answers. As such, it is practically worthless--and maybe even deceptive--as a measure of any aspect of statistical capabilities. If you take "reputation" in a wider sense as comprising all the statistics about an individual's activity, then possibly some rough sense of their capabilities can be gleaned from those. $\endgroup$
    – whuber Mod
    Nov 15, 2016 at 22:46
  • 7
    $\begingroup$ I really don't think there's much of a reliable relationship; there's some relationship because obviously it's easier for a skilled person to write an upvoted answer than an unskilled one but the relationship may be very weak -- it's possible to accumulate a large amount of reputation through answering a lot of questions while armed with mediocre skills; on the other hand a lot of highly skilled and knowledgeable people on site have low reputations, because they don't spend much time here. $\endgroup$
    – Glen_b
    Nov 15, 2016 at 22:51
  • 4
    $\begingroup$ On the other hand some of the information available can be a bit more informative, by removing the raw effect of answering a lot -- reputation per answer (though this can be a bit noisy - actual upvotes per answer will be better, if harder to get) , proportion of accepted answers, and so on. If you really want to look at total contributions rather than relative ones, choose something that's harder to get than upvotes -- number of "Good Answers" may be a semi- useful measure for example. $\endgroup$
    – Glen_b
    Nov 15, 2016 at 22:55
  • 3
    $\begingroup$ This question appears to be a duplicate of one asked four years ago by Michael Chernick at meta.stats.stackexchange.com/questions/1218. Related threads include meta.stats.stackexchange.com/questions/2107 concerning the value of CV participation in job hunting. $\endgroup$
    – whuber Mod
    Nov 15, 2016 at 22:56
  • 4
    $\begingroup$ @Glen_b Maybe--after subtracting the badges for site participation. That count is a proxy for the number of very highly upvoted posts, with a component for speedy answers. When I evaluate people on unfamiliar sites in order to ascertain who knows their stuff and can communicate it well, I look for quality and consistency as measured by relatively high votes per answer, high acceptance rate per answer, extremely few posts with zero or negative net votes, a broad range of tags (without undue specialization in any one), and more than one very highly voted post. E.g., cardinal, chl, glen_b. $\endgroup$
    – whuber Mod
    Nov 15, 2016 at 23:02
  • 4
    $\begingroup$ Some have tried. There's too much noise: you have to factor in degree of activity, time elapsed since joining, effects of all non-answer forms of reputation, and so on. You would also have to control for the thousands of different tags (it's easier to get reputation in some tags and harder in others), the time of day a question was answered, how many people were active on the site shortly thereafter, whether a post was promoted, etc. Even then you wouldn't be measuring statistical skill, but some combination of knowledge, ability to write and illustrate, perseverance, and luck. $\endgroup$
    – whuber Mod
    Nov 15, 2016 at 23:49
  • 3
    $\begingroup$ The point I was trying to get at before was that I don't think "answering a lot" (and things that depend on that most heavily) indicates anything of much value. Relative measures like reputation per answer (especially if they relate to more than a handful of answers) is more of an indication. I agree with the criteria whuber indicates. But it's hard to compare newer users with people who have been answering for years because it can take a while to learn what works for the site -- even if you're used to answering questions in other fora. $\endgroup$
    – Glen_b
    Nov 15, 2016 at 23:57
  • 6
    $\begingroup$ The short answer is evidently very weak at best. In addition to points excellently made, I add: please don't underestimate what reputation measures any more than you should overestimate it. Some reputation here is earned with much difficulty trying to help confused people asking not very good questions. Those people often don't even accept answers; the answers may not be general or good enough to gain many upvotes; they may even be pedestrian by the highest standards because they just give banal or basic advice. But the effort is all intended to help, if only the people asking the questions. $\endgroup$
    – Nick Cox
    Nov 16, 2016 at 17:52
  • 5
    $\begingroup$ Hinting darkly that many reputations are inflated and not to be taken that seriously isn't positive. It's true, but it's also unduly disparaging to those who just want to help build and maintain the site. $\endgroup$
    – Nick Cox
    Nov 16, 2016 at 17:54
  • 3
    $\begingroup$ @Carl It's difficult to say what value anybody outside CV places on CV reputation. I can believe anything from a record of posting lucid and penetrating answers helping a job application a little to it being regarded totally negatively in some universities, where only papers in top journals and/or top conferences are of interest (and any dissipation of energy elsewhere is a distraction at best). My first approximation is that it's a game you play because you find it challenging and rewarding (as many do with chess, bridge, sudoku, ....) and hope that some others benefit too. $\endgroup$
    – Nick Cox
    Nov 16, 2016 at 23:08
  • 3
    $\begingroup$ Simply going by the fact answering a popular question is much more worthy than answering an actually challenging question goes miles towards understanding what reputation reflects. It's a sign that a person can answer questions that helps the community most. In StackOverflow this is more blatant, often the highest ranked answers on a programming language are utterly basic, but that's where the community concentrates, and these are the questions that help the most members. $\endgroup$
    – Firebug
    Nov 19, 2016 at 16:23
  • 7
    $\begingroup$ One other point that seems to be ignored in the discussion above (except by Firebug recently) is that some areas of statistics and some types of questions are more popular than others. If you happen to be knowledgeable in these popular areas, you will have high reputation without a need for much statistical knowledge. So reputation to a large degree indicates how well in tune one is with the popular fields and question types rather than how good one is in statistics or in explaining statistics. $\endgroup$ Nov 20, 2016 at 10:07
  • 4
    $\begingroup$ And then there are users here with high reputation that know essentially only one thing and just keep promoting it again and again in all their answers (regardless of how relevant or irrelevant they are), and somehow (to my surprise) they still get rewarded relatively nicely (perhaps mainly by new users, but I do not have the data to support this assertion). In conclusion, I would say that no simple measure like reputations points, badges, points per answer, proportion of accepted answers etc. would indicate statistical prowess with much accuracy. All they show is how well you "fit in" at CV. $\endgroup$ Nov 20, 2016 at 10:07
  • 4
    $\begingroup$ I think @RichardHardy really means what he says; the phenomenon for a just a very small number of users is far more pronounced than a question of specialisation, but to make the point concrete would be indiscreet! There are examples of users supplying essentially the same answer again and again. If it is really off target the community does deal with it. $\endgroup$
    – Nick Cox
    Nov 20, 2016 at 17:31
  • 3
    $\begingroup$ Sure, but for every crank who turned out to be Einstein (or Galileo), thousands more really were cranks -- or monomaniacs -- in this case hammer fans to whom everything is a nail. $\endgroup$
    – Nick Cox
    Nov 20, 2016 at 19:05

1 Answer 1

4
$\begingroup$

Looking at the 2016 reputation league page:

When your fellow users vote up your questions and answers on a Stack Exchange site, you generate reputation. Reputation is a rough measure of:

  • how much the community trusts you
  • your communication skills
  • the quality and relevancy of your questions and answers

These friendly reputation leagues are an informal way of tracking your reputation within the community on a particular Stack Exchange.

Reputation is capped at 200 per day, but remember that bounty awards and accepted answers are immune to this daily reputation cap.

And I concur with that. As I pointed in the comments:

Simply going by the fact answering a popular question is much more worthy than answering an actually challenging question goes miles towards understanding what reputation reflects. It's a sign that a person can answer questions that helps the community most. In StackOverflow this is more blatant, often the highest ranked answers on a programming language are utterly basic, but that's where the community concentrates, and these are the questions that help the most members

Tag badges sound more like measures of prowess in a specific subject to me.

$\endgroup$
3
  • 1
    $\begingroup$ Different tags are differently popular. In addition, the typical amount of upvoting per view seems to differ strongly between tags. One might try to compare 2 users w/i the same tag while controlling for the number of views the threads have had since their answer was posted, but this would still measure a mishmash of things (eg, statistical insight, English proficiency, etc). $\endgroup$ Dec 8, 2016 at 16:46
  • 1
    $\begingroup$ @gung True, I didn't think about the popularity of tags (and how they come and go in questions, especially here in CV). $\endgroup$
    – Firebug
    Dec 8, 2016 at 16:52
  • 4
    $\begingroup$ You can fit a two-way random effect model with tags and people, if you like. $\endgroup$
    – StasK
    Dec 9, 2016 at 21:59

You must log in to answer this question.

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