This article develops a framework for leveraging network information about nodes over time and over multiple types of networks in order to more effectively group them into communities. The image applies this framework to congressional voting records. It reveals a congress that has been consistently split into roughly two communities, but also shows that major events have disrupted those communities periodically, leading to realignments.
Peter Mucha, Thomas Richardson, Kevin Macon, Mason Porter, Jukka-Pekka Onnela
This project explores the reliability of RDS estimates of variance, using random walk simulations of facebook networks data. The results suggest that RDS variance estimates do not accurately indicate when the true variance is high. The image presents two networks with identical degree distribution and proportions of cross-group ties that yield differing simulation results, but similar RDS estimates.
Ashton Verdery, Ted Mouw
This analysis uses a spatial process CAR model on a network defined by regional centroids with links between adjacent regions. The red regions are areas in which the relative risk of certain kinds of drug abuse is high; the blue regions are areas where the relative risk is low.
David Banks
This article examines a team recognition program at a large food company. It finds that finalists and non-finalists did not differ in terms of project resources, management or changes. However, finalist teams consistently made more connections with experts outside of the core team -- combining the benefits of a stable core group with the advantages of drawing contribution from a wider expertise network.
Jonathon Cummings
This feature requires javascript.
Duke University