ERGM

Instructor: Andy Pilny

ERGMs (exponential random graph models) are statistical models that predict the presence/absence of ties. Essentially, they view an observed network as the outcome of a set of forces that make specific micro-configurations (e.g., a closed triad) more or less likely. The main point of ERGM is to find out exactly which forces are responsible for why the network looks the way it does.

Participants will gain an intuitive understanding of why, and when, ERGMs are appropriate tools to model social networks, what ERGMs can do, and what they cannot do. The mini module will explain the theoretical foundations of ERG modeling, introduce dependence assumptions and take participants through a hands-on tutorial of running and interpreting an ERGM in MPNet. Maximum likelihood coefficient interpretation, model degeneracy, and goodness of fit will be emphasized in the tutorial.

Slides:

https://drive.google.com/open?id=1mrnKvcwZcAhl_-irPzlRvN1tsrcsMSFQ

MPNet download:

http://www.melnet.org.au/pnet/

R download:

https://cran.r-project.org/mirrors.html

Supplementary resources:

The best non-technical introduction to ERGM I can find:

Valente, T. (2010). ERGM. In Social networks and health: Models, methods, and applications. Oxford.

Brief non-technical introduction and extension to valued networks:

Pilny, A., & Atouba, Y. (2018). Modeling valued organizational communication networks using exponential random graph models. Management Communication Quarterly, 32(2), 250-264. doi: 10.1177/0893318917737179