26
$\begingroup$

I thought MO users might be interested in this graph constructed by Neal Young in a discussion at meta.cstheory about the more pronounced downward trend in cstheory. It shows MO Q&A activity from Aug 2013 to Feb 2016 (he produced similar graphs for cstheory and cs, available at the above link).


          DownwardTrend


$\endgroup$
12
  • 1
    $\begingroup$ ...and is this because math.se is going up? Or is it also going down? $\endgroup$ May 12, 2016 at 14:00
  • 2
    $\begingroup$ @GeraldEdgar: I don't know. If one were a master of the data facilities provided by SE (which I am not), your question could be answered. $\endgroup$ May 12, 2016 at 15:26
  • 2
    $\begingroup$ @GeraldEdgar: math.se has a much higher volume of Q&A with <5 votes. However, there is a slight decline in voted Q&A: data.stackexchange.com/math/query/485296/… $\endgroup$ May 12, 2016 at 16:12
  • 4
    $\begingroup$ @FrançoisG.Dorais Of course, an older Q/A has also had a longer time to accumulate 5/10 upvotes. I've adapted that query to only count upvotes cast within a certain user-specified window of post creation, and it seems much more flat (I've only looked at 7 day and 30 day windows). Query link. $\endgroup$
    – user642796
    May 12, 2016 at 18:00
  • 2
    $\begingroup$ Joseph, could you state more clearly whether the graph in your post shows information about MathOverflow or information about cstheory? $\endgroup$ May 12, 2016 at 18:11
  • 1
    $\begingroup$ @arjafi: MO is also mostly flat with a window. $\endgroup$ May 12, 2016 at 18:35
  • 2
    $\begingroup$ @JoelDavidHamkins: Thanks, I didn't notice the ambiguity until you pointed it out. The graph shows MO activity. $\endgroup$ May 12, 2016 at 19:19
  • 1
    $\begingroup$ Here is a link to the Data Explorer query that produced Neal Young's data: query link. $\endgroup$ May 12, 2016 at 19:45
  • 5
    $\begingroup$ Would be interesting to plot this against the dynamics of the number of active users too, if such a thing is available. $\endgroup$ May 15, 2016 at 20:03
  • 3
    $\begingroup$ The blue does not look like a downward trend from 2014. Regression lines are often misleading. $\endgroup$
    – Kimball
    May 21, 2016 at 0:06
  • $\begingroup$ @Kimball Why not? It looks downward to me. $\endgroup$ May 24, 2016 at 2:44
  • $\begingroup$ I will add that some statistics about questions, answers, votes and traffic can also be found in site analytics. However, this is only accessible to 25k+ users. For example, some stats about traffic are shown in this answer: The number of on line users at a given moment. $\endgroup$ Dec 4, 2017 at 11:31

2 Answers 2

11
$\begingroup$

I have one theory that may explain this trend. Specifically most of the questions I have wanted to ask have already been asked. I think this is a good thing in many respects. Not so good for the community, but great for learning. Namely you will see the number of questions new to be proportional to the number of users times the inverse of the probability of finding the answer.

$\endgroup$
1
  • 4
    $\begingroup$ +1. And now the answer that I have wanted to give has already been given :) $\endgroup$ May 26, 2016 at 15:08
9
$\begingroup$

Might be offtopic, but again maybe not. In any case this is not really answering the question, just some additional information (seemingly) related. I've rerun the query about total questions and answers together for the period indicated (could not run it for longer time stretch). What I've got is this:

enter image description here

Seems like for the first time in 2016 the total number of questions has exceeded the total number of answers.

I believe one might conclude that the number of questions actually grows, although more slowly than the number of answers drops.

And here is an update for 2015 - 2017. I believe it rather confirms than contradicts the suggestion that there is a tendency for a combination of an upward trend in questions and downward trend in answers...

enter image description here

...And now one more update. I dare claim that the trend is more and more apparent! Can anybody tell me how to evaluate actual statistical significance of it?

enter image description here

$\endgroup$
2
  • 14
    $\begingroup$ Just a note that the spike in the number of questions in the last "month" is likely due to a number of closed and not-yet-deleted questions. With the noted reticence of MO users to manually delete questions, such questions will stick around until the roomba clears them out (often about nine days after closure, though some criteria may push this to 30 days). $\endgroup$
    – user642796
    May 16, 2016 at 17:27
  • 1
    $\begingroup$ @user642796 seems like that spike was genuine. It is still not enough evidence to be confident that questions go up while answers go down... $\endgroup$ Dec 4, 2017 at 16:21

You must log in to answer this question.

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