6

I have an API connection to stack exchange (and implicitly the Mathematica stack exchange) which I use in day-to-day Mathematica usage, but which I've recently been using for things like calculating the average user reputation (500 for users with >100 reputation, 176 for absolutely everyone, for those who were wondering) and other fun stuff like that.

I was wondering if there's an official collection of stats like that so I can check mine against those to see what sorts of things my API connection misses. Is there such a collection?

Edit

This is admittedly off topic, but interesting:

Here's the distribution of user reputation as of 1/6/2017 (with 101 mapped to 1 because the vast majority of people with reputation 101 have it due to the association bonus (trust me, I plotted it out and there's a huge plateau at exactly 101)):

enter image description here

And here's the same thing log-plotted so Szabolcs and Mr.Wizard skew the distribution a tiny bit less:

enter image description here

And here's the PDF plot of the estimated `WeibullDistribution of user reputations (chosen only because the docs had a picture that looked kinda how I expected the distribution to look):

enter image description here

To quote march in the comments: "that's what we call a skewed distribution!"

Update 2

I compressed all the user data (It's missing the last 325 users, but they all have rep 1 and so don't affect the statistics too much) and have dumped it here. If that file ever goes down just shoot me a message and I'll re-upload it.

  • What is the median user reputation? – march Jan 6 '17 at 18:44
  • 101 across the first 10,000 users. I'll check across a larger set, too (assuming I don't hit the API call limit). – b3m2a1 Jan 6 '17 at 18:48
  • 101! Exactly the number that a user gets if they have enough reputation on another site and join this one without gaining any rep whatsoever. That suggests that more than half of the users are essentially one-time users or not users at all, and that there are high-reputation outliers that are way above the median (and, of course there are). – march Jan 6 '17 at 18:51
  • Yep. If you want I'll plot the data in a moment so we can see exactly what this looks like. I can also sort and normalize by user id, because observationally it seems the users who have been here longer have higher reputation. – b3m2a1 Jan 6 '17 at 18:53
  • Interestingly, across all users it's 13 ( I say all because I asked for the first 50000 users and only got 29825). And that plot has a strong plateau at 101. So clearly the median should be lower but we have a lot of cross-over. (in fact it's 1 if we send 101->1) – b3m2a1 Jan 6 '17 at 18:57
  • The median is one! That's what we call a skewed distribution! – march Jan 6 '17 at 21:56
  • It's pretty wild. I'll edit the question with some fun plots. – b3m2a1 Jan 6 '17 at 21:58
  • @march you might enjoy the plots. They definitely confirm the skew. – b3m2a1 Jan 6 '17 at 22:16
  • 1
    I wonder if Szabolcs and Mr. Wizard feel guilty that they skew the dist on the 1st plot so much. – corey979 Jan 7 '17 at 17:41
  • @corey979 from the log plot we can see that Mr.Wizard has more reputation than there are 1-and-done users of the site. (we can't say he has more reputation than them put together, as many have 101 rep, but it's still a fun statistic). – b3m2a1 Jan 7 '17 at 17:49
  • @corey979 Another amusing statistic (maybe I should just do a discussion post on this, given how much fun I'm having doing this) is that the MSE reputation Gini coefficient is .88, using the method here. That means we are a hugely unequal nation (US Gini is ~.4 for comparison). – b3m2a1 Jan 7 '17 at 17:57
7

The StackExchange Data Explorer has a large collection of queries that you can run and compare your own results with.

For example this query shows that the average reputation of users with more than 100 reputation is 490, which agrees well with the number you found.

  • I've been wondering how can I know how close I am to the Pundit badge, and there it is (not even half way there, btw ;) Thanks! Do you know if I can check which of my comments were the highest voted ones? – corey979 Dec 31 '16 at 18:48
  • @corey979 Sure, I typed "comments" into the search box and got lots of queries that'll do that. – C. E. Dec 31 '16 at 20:24
  • This is super cool. They don't happen to have an API for this do they? – b3m2a1 Jan 2 '17 at 6:16
  • @MB1965 No, but actually StackExchange very recently made data available through Google BigQuery, which also allows you to use SQL syntax, and it does have a REST API. – C. E. Jan 2 '17 at 6:23

You must log in to answer this question.

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