Sunday, November 07, 2010

Follow Up: Twitter BFFs

Last time I tried to do this, I tried to use as large a range as possible for the data and found some people missing. I still don't know why this is. But coming back to it, I tried looking at just the previous week, and the missing people reappeared.

I took the numbers for the last two weeks. I don't entirely trust them, but as always I try to work with what we can get.

[edit] - I just checked the numbers again, over the same period, and the numbers for follower replies have changed fairly significantly. This displeases me.

The Numbers!

Using the method outlined in the previous post, the fifth column is the geometric mean of mentions (assuming a two-way connections between myself and all listed). For comparison, the fourth column is the arithmetic mean. The numbers are fairly similar.

As I said last time, I can't adjust for talkativeness. It's possible the picture would be dramatically different if I did have the numbers, but I guess we'll never know.

The sixth column is just the geometric means as a percent proportion of the total sum of the geometric means.This is what the names have been sorted by (in descending order).

[I'd make a pie chart, but frankly, I can't stand them.]

[Side Note] - the correlation coefficient of the replies is 0.62 (on a scale of 0 to 1).


So there you have it. You can take the first three or five names and call them my top friends (or BFFs, if you're so inclined) for the last two weeks.

Connection Graphs

Firstly, here's a table of how many friends I have in common with friends.
As I said last time, you could incorporate this into the "BFF score". I didn't, but it's there in case you're interested.

Aside from that, here are a couple of coloured network graphs.

First, men in blue, women is red. No immediately obvious patterns.
Second, people I know in real life in red, web people in blue.
This one's slightly more interesting in that the IRL friends are all clustered together in the middle with all the web friends spread around the edge. Which you might well have expected, but it's nice to be able to see it.



No comments: