An association chart, and the arithmetic underneath it LINK
A chart drawn by hand shows what you already believed. The value is in the measures that find the person holding two groups together when nobody can see it.
The links
A CSV with two columns saying who is connected to whom. Call data, a visitor log, a list of directorships, co-defendants, people seen together: anything that is a list of pairs. It is read in this tab and goes nowhere.
Which column is which
The network
The chart
What the numbers say
Everybody, by betweenness
How two of them are connected
The shortest chain between any two names in the file, and whether there is more than one chain that short.
Take it with you
How it works
LINK turns a two-column list of who is connected to whom into an association chart, and does the arithmetic a hand-drawn chart cannot. Betweenness finds the person who holds two groups together, who often matters more than the person with the most links. The layout is the same every time for the same data, so charts can be compared. It handles large networks, and nothing is sent anywhere.
The guide: Association charts, and the broker
Questions people ask about LINK
What does the file need to look like?
Two columns of names or identifiers, one pair per row: who and who. Anything else is optional (a column saying what kind of link it was, a count or weight, a date) and is carried onto the chart where it exists. If the headings are recognisable they are matched up automatically, and if they are not, the first two columns are used, because a two-column file with odd headings is still a list of pairs.
What is betweenness, and why does it matter more than the number of links?
Betweenness counts, for every pair of other people in the network, how many of the shortest routes between them pass through a given person. Somebody connected to everybody has a high number because they are a switchboard, which is obvious and rarely useful. Somebody with two or three links and a high number is sitting between groups that have no other way of reaching each other, and that is the finding: remove them and the network does not just lose a node, it comes apart. The page ranks people by the ratio of betweenness to links for exactly that reason.
Why is the chart the same every time?
Because a force-directed layout is random unless you make it otherwise, and a random layout is useless as evidence. The simulation is seeded from the names in the file and runs a fixed number of rounds, so the same data always produces the same picture, which means you can put it in a report, compare it against one made a month ago, and have somebody else reproduce it.
Can it handle a big network?
Comfortably into the low thousands of people. The measures themselves are the expensive part: betweenness costs a breadth-first pass for every person, which is the honest cost of the exact answer rather than an estimate. The layout runs fewer rounds on a large network, and the chart lets you hide anybody below a number of links so a hairball becomes readable, while the measures below stay computed on the whole thing.
What about direction: who called whom?
Links are treated as undirected: a call from A to B and a call from B to A are one link between A and B, with the rows counted. That is the right model for association, which is about who is connected to whom rather than who initiated it. If direction is the question you are asking, the tool for it is the one that reads a schedule of call data, which keeps the direction of every event.
Is any of it sent anywhere?
No. The file is read by the browser, every measure is computed in the page, and the SVG, the analysis and the CSV are all written by the tab. There is no account and no server, and the file is not kept across a reload.
Doing this often? OBSCURA Case maps the people in a collection and who wrote to whom, beside the chronology and exhibits. Everything is US$5 a month or US$48 a year.