researcher2 ([info]researcher2) wrote in [info]lj_research,
@ 2005-01-31 13:48:00
Previous Entry  Add to memories!  Tell a Friend!  Next Entry
Some statistics
I have a copy of the friends graph from a month or two ago. It's 842M uncompressed, 228M with gzip.

Some statistics:

1. Mean number of friends is 13. (More precisely, 12.95.)

2. Median number of friends is 5.
(The numbers are the same measuring "friends-of". That was obvious in the first case, not so obvious in the second.)

For friends: 10% have no more than 0, 20% have no more than 1, 30% have no more than 2, 40% have no more than 3, 50% have no more than 5, 60% have no more than 7, 70% have no more than 11, 80% have no more than 18, and 90% have no more than 33.
The friends-of numbers are: 1,1,2,3,5,7,17,32. I don't have combined numbers, though that might be an interesting statistic.

Here's some numbers (my total data, collected by crawling the LJ friends graph over the course of a month, had 2696203 people in it):
# with 0 friends: 299533
# with 1 friend: 414359
# with 2 friends: 263776
# with 3 friends: 193527
# with 4 friends: 155717
# with 5 friends: 129682
(For friends-of, the numbers are: 144064, 517753, 307090, 210998, 159798, 128749 respectively.)

The distribution of both of these is skewed. Here's a histogram using a log-log scale. I'm not used to thinking in log-log, but the graph is ugly and senseless without the log-log since it goes for the extremes. But here's a limited shot between twenty and one hundred.

Now, here's the big graph, a plot of friends-of (indegree) verses friends (outdegree). (I warn you, it is big. It has over two million points and takes a while to download and draw.) Here you see something interesting that sorta comes out in the other two graphs: there are more people with lots of friends-of than people with lots of friends.

Who are the extremely popular folk? Here's a list of the people with over 2500 on their friends-of list:

</a></b></a>[info]doctor_livsy at 4601
</a></b></a>[info]quizdiva at 4590
</a></b></a>[info]dimkin at 4554
</a></b></a>[info]status at 3961
</a></b></a>[info]kim_jong_il__ at 3438
</a></b></a>[info]dolboeb at 3142
</a></b></a>[info]avva at 2870
</a></b></a>[info]thegraybook at 2784
</a></b></a>[info]mistersleepless at 2630
</a></b></a>[info]teh_indy at 2535
</a></b></a>[info]drugoi at 2591

</a></b></a>[info]foma and </a></b></a>[info]fif are very friendly, with 1954 and 2219 friends, respectively. (I kept track of those with more than 750 friends. I think there may have been a policy change at LJ during my data collection, since there's a sharp cutoff in my data around 750.)



p.s. My data is out-of-date, but it probably is reasonably okay when it comes to trends. Also, while I did sanity check these results a bit, I won't swear they are correct.



(Post a new comment)


[info]nibot
2005-01-31 03:55 pm UTC (link)
I'd like to get a copy of the friends graph... is this something I could get from you?

(Reply to this)


[info]endquote
2005-01-31 07:07 pm UTC (link)
I dunno, I'm pretty sure I have way more than 13 mean friends.

(Sorry, had to...)

(Reply to this)


[info]evan
2005-01-31 07:36 pm UTC (link)
http://www.livejournal.com/~evan_tech/90013.html

we put a cap on the number of friends at one point; those two must've done it before the cap was made.

(Reply to this) (Thread)


[info]researcher2
2005-02-01 05:38 am UTC (link)
Thanks. I was wondering why there seemed to be such a sharp cut off. I suspected it was a policy.

(Reply to this) (Parent)


[info]bookshop
2005-09-01 07:20 pm UTC (link)

Are you guys ever going to consider lifting the friends limit???

(Reply to this) (Parent)


[info]pinkfinity
2006-05-08 12:22 am UTC (link)
Is there ever going to be a lift of the friends limit? I keep hitting the 750, and would love to be able to add a bunch more.

(Reply to this) (Parent)


[info]datawar
2005-02-01 06:54 pm UTC (link)
Awesome. So this is the strongly connected component? Any idea what lies outside?

(Reply to this) (Thread)


[info]researcher2
2005-02-01 07:07 pm UTC (link)
This is, at the time of my crawl the weakly connected component.

I haven't done the calculations to find the strongly connected component.

I do have plans to estimate the users outside the giant connected component using data I've collected from the lastest post page, but that would only be a crude estimate, and I don't know if I'll have time.

(Reply to this) (Parent)


[info]tasha
2006-02-15 12:00 am UTC (link)
Are you going to do any more research similar to this, with maybe results for most popular communities? I'm highly interested in knowing this, and wonder why it's not already a feature of sorts, so you can automatically see some of the best/most popular communities out there. Thanks. :)

(Reply to this) (Thread)


[info]researcher2
2006-02-15 01:27 am UTC (link)
Unfortunately, no. I know other people have data on communities, but I'm not collecting any.

(Reply to this) (Parent)


Create an Account
Forgot your login or password?
Login w/ OpenID
English • Español • Deutsch • Русский…