Graphical discovery in stochastic actor-oriented models for social network analysis
Is Version Of
This work presented in this thesis combines statistical models for social networks and network visualization in new and exciting ways. In Chapter 1, a thorough review of the literature in the topics of statistical network models and network visualization is presented. In Chapter 2, we focus in on one type of model for dynamic social networks: the stochas- tic actor-oriented models (SAOMs), introduced by Snijders (1996). Unlike other network models, SAOMs are not very well understood, so we use model visualization techniques in- spired by those introduced in Wickham et al (2015) in order to make the models a little less murky. The SAOMs are a prime example of a set of models that can benefit greatly from application of model visualization, and with the help of static and dynamic visualizations, we bring the hidden model fitting processes into the foreground, eventually leading to a better understanding and higher accessibility of stochastic actor-oriented models for social network analysts. In Chapter 3, we further explore the SAOMs using the visual inference methodology of Buja et al. (2009). We construct significance tests of model parameters, goodness-of-fit tests, and power calculations for the objective function parameters in SAOMs using visual inference. In this way, we can explore complex network data more completely than traditional significance and goodness-of-fit methods that rely on one-dimensional de- rived features of networks do. In Chapter 4, we present an R package for drawing networks using the popular grammar of graphics R plotting paradigm, ggplot2 (Wickham 2016). We close with a discussion of the limitations of the work and directions for the future in Chapter 5.