A matrix of Astronomers

Here I show a matrix representing the connections between users; it shows the same data as the constellation of astronomers (aka hair ball), but in a different format. The current options for ordering users along the axes are not well matched to the structure of the graph, but they provide some insight into the distribution.

The page is updated infrequently.

Time range (Pacific): to

The vertical axis represents accounts that made a tweet, and the horizontal is for those accounts that are mentioned in a tweet. The circle size indicates the number of times that the user was mentioned or retweeted (this can be changed using the buttons above the figure). Vertical lines represent accounts that were mentioned a lot (since the data includes retweets then the default viewis dominated by the re-tweets, but you can now turn hide these) and horizontal lines mean that that account mentioned a lot of other accounts.

The color of each circle is an indication of the "community" to which the connection belongs. This is experimental and is likely to change (e.g. the mapping between community and color). The colors are the same as used in the network view (aka hairball) of the same data. I used the algorithm of Ahn, Bagrow, and Lehman, "Link communities reveal multiscale complexity in networks", Nature (doi:10.1038/nature09182) - see their project page for more information - which groups together links into communities (rather than the more common approach of grouping nodes, so in this case users, into communities). Early experiments using Gephi suggest that this is going to take some experimentation to get anything usable. Note that the algorithm I implemented assumes that the graph is undirected - i.e. it does not account for the fact that just user1 mentioning user2 does not imply that user2 has mentioned user1 - and

Hovering over a circle will give details on the connection and will enhance other connections in the same community (and fade out those that are not).

The ordering of the users on the axes can be changed using the menu above the figure. I am exploring choices to bring out the structure of the community, but so far it's a bit messy (which is likely a reflection of the "truth"):

by user number

This is the default value and you can consider the position to be a random value; it is actually related to the numerical identifier assigned by Twitter, so values on the left (or top) of the axis are likely to have been on Twitter longer than those to the right (bottom), but this is not guaranteed and may change at some time.

by number of tweets

The users are ordered by the number of tweets made about the conference. The axes are actually in descending order, so the top/left position represents the most tweets and the bottom/right the least tweets. In this case the number of tweets includes retweets.

by number of followers

The order is by the maximum number of followers the user had during the conference (this is actually approximate since I only got this value whenever the user made a tweet or was retweeted). The top/left position is for the user with the largest number of followers, and the bottom/right is for the smallest number of followers.

by number of followers that also tweeted about AAS225

This is the same as the number of followers except that we are only counting those followers of a user that also tweeted about AAS 225. This is a very simplistic attempt to distinguish between "Astronomers" and "people who follow accounts like @SPACEdotcom or @BadAstronomer and retweeted one of their messages". Really I should be doing some sort of community detection and use these communities as an ordering, but I have not got around to this yet.

by number of friends that also tweeted about AAS225

This is the number of people that the user follows that also tweeted about AAS 225.

The diagonal black line is a guide, and indicates the position of accounts who mentioned themselves.

Please see the data collection page for a description of how the data was collected and processed.

Credits

This visualization was created using the d3.js JavaScript library.