Parts | Days | Selection | Search | Updates | Downloads | Help
SOE: Fachverband Physik sozio-ökonomischer Systeme
SOE 6: Dynamics of Social and Adaptive Networks I
SOE 6.1: Talk
Thursday, September 30, 2021, 11:45–12:15, H3
Understanding force directed layouts through latent space models — •Felix Gaisbauer, Armin Pournaki, Sven Banisch, and Eckehard Olbrich — Max Planck Institute for Mathematics in the Sciences, Leipzig, Germany
This contribution brings together two strands of research: Latent space approaches to network analysis and force-directed layout algorithms. The former can be considered as extensions of spatial random graph models for social networks, which have the goal of embedding a graph/network in an underlying social space [1] and have been employed successfully in the estimation of ideological positions from follower networks on Twitter [2]. The latter are used ubiquitously for data exploration, illustration, and analysis. Nevertheless, an interpretation of the outcomes of graph drawings with force-directed algorithms is not straightforward. We show that interpretability can be provided by random graph models in which the nodes are positioned in a latent space. The closer the positions of the nodes, the more probable it is that they are connected. We show that force-directed layout algorithms can be considered as maximum likelihood estimators of such models. We also present ready-to-use implementation of the layout algorithm and show its application to Twitter retweet networks.
[1] P. D. Hoff, A. E. Raftery, and M. S. Handcock (2002). Latent space approaches to social network analysis. Journal of the American Statistical association, 97(460), 1090-1098. [2] P. Barberá (2015). Birds of the same feather tweet together: Bayesian ideal point estimation using Twitter data. Political analysis, 23(1), 76-91.