If you work with longitudinal network data and will be at the 2016 INSNA social network analysis conference in Newport Beach USA, I hope you will consider attending our workshop on April 5th:
First, a confession: I never read Lord of the Flies. I remember thinking as a kid that it sounded scary and grim, and have yet to get around to reading it as an adult. But now I don’t have to because, thanks to Jimi Adams, I can watch a network animation of the the evolving social structure in the novel :-)
Jimi used the book in an undergrad SNA course as a way to illustrate network concepts and metrics, reporting the experience in a paper Using Lord of the Flies to Teach Social Networks. In the process he coded up all the conversations between characters in the novel (directed blue ties) — as well as their co-appearances in the same locations (undirected gray ties) — as a longitudinal network. The 473 time slices were then rendered as a video using the ndtv R package. The paper gives links to time points in the movie where the network structures can be used as examples of the concepts he discusses.
Jimi made the data available for download as a single R networkDynamic object at http://goo.gl/YWMSXQ. Encoded in the object are the durations accounting for node (dis-)appearances, durations for speaking and co-presence ties, as well as dynamic node attributes to indicate possession and transfer of the conch (the boys’ ceremonial ‘talking stick’, shown in green), and deaths. From correspondence, I’ve learned that the few red ties indicate violent conflict.
I was planning to post here an alternate temporal visualization using just one of the tie types, but realized that I’ll need to look at this in more detail in a future post. I had been thinking of this network as multiplex (with the potential for more than one edge active between a pair of vertices) but in fact tie types are coded as a dynamic ‘weight’ attribute with with discrete values for each type. This works fine for the movie but it means that tie types are mutually exclusive. Although pair of characters can’t be coded as simultaneously co-present and talking, as a work around we could probably infer the former from the latter since conversations are face to face. However the main issues is that, off the top of my head, I don’t have a method for extracting one dynamic network from another including edges as active only when an attribute is active with a certain value. Stay tuned ..