7

To quote from mSO, since my request is pretty much identical:

I would like to have statistics for the quality of posts coming from unregistered versus registered accounts on my site.

It is my suspicion that people who are willing to put out the effort to register an account will also put more effort into writing good questions; I'd like to see some evidence for or against this.

I also suspect that on my site the vast majority of good questions and answers come from registered users, but again I'd like to see some data.

Is this already possible through the Data Explorer? If so, how? If not consider this a feature request.

The reason for this is I am considering proposing that Server Fault stop accepting questions from unregistered users, and I would like to have statistics to analyze as part of that process.

  • 1
    Now that we have the stats, how do you intend on using them? Is it based on volume of bad questions, or the percentage? 256 bad questions in 2 months isn't so bad compared to 2,181 bad questions from registered users. That's (at a very rough estimate) about 1/5th of bad questions are from unregistered users. And then what about the 43% that weren't bad questions? – Mark Henderson Sep 13 '13 at 5:08
  • I'm still thinking about it. The stats are... definitely not quite what I expected. – Michael Hampton Sep 13 '13 at 5:09
  • 3
    The majority of people who ask questions on SF have accounts on SO so the results below aren't surprising as SO users have to be registered and we know from revious experience they ask crappy questions too. – Iain Sep 13 '13 at 7:29
  • 1
    Another perspective: If only 448 questions (6%) and 379 answers (5%) have come from unregistered users, then why keep the unregistered route at all? It is pretty clear that registration is not a big barrier to entry in this case. Normally, guest posting, guest reviews and guest checkout etc are used so as to not turn away large proportions of participants (often bigger proportions than registered folks) but in case of SF, that's clearly not the case - most people register anyway. – OC2PS Sep 13 '13 at 11:38
  • 1
    Much, much more importantly, 32% of registered questions being bad is a real slam. Perhaps we are too rigid in closing/deleting questions? Or maybe we don't clarify our guidelines prominently enough? (we ARE pretty different from other SE sites) – OC2PS Sep 13 '13 at 11:43
  • @OC2PS: I suspect that the numbers would be higher if we had more people who voted to close. As it is we have <300 and <50 of them use the privilege with any regularity. – Iain Sep 16 '13 at 16:46
11

As usual, I'm gonna define a "bad" post as one that has been closed, deleted (by someone other than the author), or down-voted below zero. In the past 60 days,

Questions

Registered Unregistered Registered Badness Unreg Badness PctRegBad PctUnregBad 
---------- ------------ ------------------ ------------- --------- ----------- 
6853       448          2181               256           31.83     57.14       

Answers

Registered Unregistered Registered Badness Unreg Badness PctRegBad PctUnregBad 
---------- ------------ ------------------ ------------- --------- ----------- 
6995       379          704                188           10.06     49.6        
  • 2
    Jesus if down-voted below zero and moderators deleting questions define "bad" then how to you define "censored"? – Evan Carroll Sep 13 '13 at 23:52
  • 12
    Blocked automatically by the quality filter, of course. Also, I'm not Jesus. – Shog9 Sep 13 '13 at 23:54
  • Right, and that's obvious. Jesus would not sit back idle as a faulty quality filter censored my questions. – Evan Carroll Sep 14 '13 at 0:44
  • 11
    I don't think the fault lies in the quality filter. – Falcon Momot Sep 14 '13 at 4:34
3

An interesting conclusion that can be drawn from Shog9's numbers is that only 11% of bad questions are asked by unregistered users (21% of bad answers are). Removing unregistered users would cut the appearance of bad content (questions and answers both) on the site by 13%, assuming none of them registered.

Also, disposing of all those people would have deprived us of 4% of our good questions and 3% of our good answers, or 3% of our total good content. In other words, the site would shrink by up to 3%.

Note that the percentages in the first and second paragraphs are not directly comparable.

If the unregistered users created bad content with equal propensity to registered users, it would be expected that they asked 143 bad questions and 38 bad answers, or created 181 bad pieces of content in total. Using a chi-squared test to determine whether there is a statistical correlation between a lack of registration and the creation of bad content, I find P<0.0001 when considering only questions, only answers, and both. This shows that in all cases, these numbers are statistically meaningful.

  • 1
    It is perhaps not such a bad assumption that if we haven't caused the conditions of "badness" to occur (downvoting, closing, and deleting), the content is considered useful, given our collective propensity to downvote and close. – Falcon Momot Sep 13 '13 at 5:31
  • 2
    I think the spread is more significant than the absolute quantity. A 25-point spread on questions isn't terrible (it's not GOOD, but if registered users are screwing up I can't really expect total newbies to get it right). The spread on answers however... ::shudder:: ::walks off mumbling something about "bloody spammers" and "kill everyone":: – voretaq7 Sep 13 '13 at 15:34

You must log in to answer this question.

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