So far, we have mainly been doing descriptive statistics with network data. That is, seeing how we work with build network data structures and describing the properties of those networks. For example, the question “how are the degrees distributed in this network?” is a descriptive question in that we are describing the distribution of ties.
Moving forward, we will shift toward inferential statistics where we test hypotheses on networks. For example, the question “is the distribution of degrees different from a network where ties form at random?” is a hypothesis and we need a modeling framework for testing this hypothesis.
This provides a whole new body of questions that we can ask:
The lecture introduces the Exponential Random Graph Model (ERGM), which is a flexible tool for specifying network configurations that generate global network structures.
The first lab provides an introduction to ERGMs in R. We will work
through different dependence specifications of the model reviewed in the
lecture (e.g. dyadic independence) and show how to incorporate node
level attributes into the model. We will be using the ergm
package throughout the lab.
The second lab examines functions for goodness of fit and simulation
in the ergm package. A comprehensive archive of materials,
discussion, syntax, papers, etc. for the ergm package is
available at the statnet website.