SEDE queries related to average reputation

Conversation started Jul 6, 2020 at 14:05.

Martin Sleziak

Jul 6, 2020 2:05 PM

As I mentioned in a comment, you can see some kind of distribution of users by reputation in the reputation league. (Notice the table on the right. Reputation leagues is also linked from the thingy in your profile which says "top x% this week/month...") If you want to calculate average reputation...

It's interesting that once you reach a 1000 points, the average is 6000+, but over 10k we only reach 38k or so on average. I wish we could break this down by length of time since registration, number of questions, answers, bounties, etc. — Asaf Karagila ♦ 25 mins ago

@AsafKaragila As it happens, some time ago I created a query showing average reputation depending on the age of the account in months: data.stackexchange.com/math/query/1241593/… (I hope I did not make some mistake there.) If there are some other queries which can be clearly specified and which are not beyond my abilities, I might try - but I cannot promise anything for sure. (Perhaps it might be better to discuss them in chat - as not to leave too many comments here.) — Martin Sleziak 18 mins ago

@AsafKaragila I am not really sure how close this is to what you wanted, but I have tried something like Average reputation of users depending on number of posts: data.stackexchange.com/math/query/1259417/…

Although I suspect that in the higher numbers, the average is probably average of one user.

The part of the graph with reputation 1 in the middle seems strange, but here are two suspended users who have around 6300 posts: data.stackexchange.com/math/query/1259418/…

Maybe iuf we do not divide the x-axis so fine, it give a bit more information. (Or at least, there are less cases where the average is taken only from a single user.) data.stackexchange.com/math/query/1259417/…

That's enough playing with SEDE for today - I should try to do some work, too.

Martin Sleziak

Jul 6, 2020 2:51 PM

I see I have forgotten to include the second graph.

Clearly, not all users have the same average score (or reputation gain) per post - otherwise the graph would be linear.

Average score per answer, non-CW version

Similar query restricted to a single tag and a non-CW version.

Of course, I have posted some similar queries in the praise chatroom before: chat.stackexchange.com/search?q=average&room=17403

in In praise of Math.SE site and its users, Aug 7 '19 at 7:40, by Martin Sleziak

Top 500 answerers by average score: main and meta.

in In praise of Math.SE site and its users, Aug 7 '19 at 7:42, by Martin Sleziak

The same query restricted to users with at least 100 answers: main, meta.

2 hours later…

John Jiang

Jul 6, 2020 5:05 PM

@MartinSleziak hello Martin, I think I finally joined. The differential galois theory group was created a long time ago.

8 hours later…

Martin Sleziak

Jul 7, 2020 12:36 AM

@JohnJiang Yes, that's true. I have mentioned it just as an example. (It was one of the attempts to create a reading group in chat.)

Sorry for the chaos in the comments. I should have simply posted the comment as I did now - instead of suggesting to delete the comment after you see it and complicating things.

Let us hope you get some response on MO.

Maybe later I can offer bounties on the Mathematics questions (How to organize math study groups online for long-distance collaboration? and How do I find partners for study?) - just to see whether that brings some new interesting answers.

Martin Sleziak

Jul 7, 2020 12:57 AM

BTW I noticed this - relatively recent - answer:

A: How do I find partners for study?

There is this site: https://www.reddit.com/r/STEM_Study_Groups/ And this: https://my.cuddy.app/

It mentions reddit.com/r/STEM_Study_Groups and my.cuddy.app

Martin Sleziak

Jul 7, 2020 1:11 AM

Although I am not sure whether either of those links has some useful content.

1 hour later…

Martin Sleziak

Jul 7, 2020 2:18 AM

I though that looking at the total score (rather than reputation) would make the graph more linear - by removing the effect of the suspended users who have reputation 1 despite many posts.

SEDE: data.stackexchange.com/math/query/1259600/… and data.stackexchange.com/math/query/1259600/…

It only worked to some extent.

Since we're looking at users divided by number of posts, perhaps it is good to include also the list of users with most posts: data.stackexchange.com/math/query/1259606/users-with-most-posts

2 hours later…

Martin Sleziak

Jul 7, 2020 3:50 AM

I take it all these "averages" are arithmetic means. Anyone keen to calculate the mode? median? geometric mean? harmonic mean? $\ell^2$-norm? standard deviation? — Gerry Myerson 21 mins ago

@GerryMyerson I think that to do such computations it might be better not to do them directly in SQL, but to download the data and do them in some other software. I think this might be difficult to do in SQL - at least I can say for sure that it is beyond my knowledge of SQL. — Martin Sleziak 2 mins ago

I have at least tried to find median, using the method suggested in Jeff Attwood's answer: Function to Calculate Median in SQL Server. Assuming that I did not make some mistake, median is equal to 1. I guess this was expected. Again, we can restrict this to users above some threshold, for example at least 102 reputation (removing all users who only have repuataion from the association bonus.) — Martin Sleziak 2 mins ago

Milo Brandt usefully summarized the results for averages:

For anyone who doesn't feel like running the mentioned SEDE queries to get current data: as of when I posted this comment (July 6th, 2020), the average reputation was 123.5, the average of those with reputation >=2 was 269.7, reputations >=101 average 543.5, and reputation >=102 average 889.7. — Milo Brandt 5 hours ago

Similarly, by changing the threshold, I get different values of median.

min=1 - median is 1

min=2 - median is 75

min=101 - median is 111

min=102 - median is 169

min=125 - median is 249

min=500 - median is 1053

min=1000 - median is 2135

min=10000 - median is 19522

Conversation ended Jul 7, 2020 at 3:56.

all rooms

Jul '206