Speaker
Robert Krause
(University of Kentucky)
Description
Estimating exponential random graph models (ERGMs) with missing data on nodal attributes (e.g., missing gender, age, ethnicity, ...) is currently not possible with the ergm() function of the ergm R-package, the most used software to estimate ERGMs. We have developed a new estimation of Bayesian ERGMs, implemented in the bergmM() function in the Bergm R-package, capable of estimating (Bayesian) ERGMs with missing nodal (and missing tie) data. The algorithm also provides, if desired, multiple imputation of the attributes (and tie-variables) to be used for additional analyses.
In this presentation, we will explain the algorithm, its limitations and assumptions, as well as how to use it in practice.
| Topics | • Statistical methods and models for network analysis |
|---|---|
| Keywords | missing data, ergm, bergm |
Primary authors
Robert Krause
(University of Kentucky)
Prof.
Nynke Niezink
(Carnegie Mellon University)
