
CMUISR10130
Institute for Software Research
School of Computer Science, Carnegie Mellon University
CMUISR10130
The LInk Probability Model: A Network
Simulation Alternative to the Exponential
Random Graph Model
Ian McCulloh*, Joshua Lospinoso**, Kathleen M. Carley
December 2010
CMUISR10130.pdf
Center for the Computational Analysis of Social and Organizational Systems
CASOS Technical Report
Keywords: Exponential random graph models, p* models,
statistical models for social networks, degeneracy, longitudinal social
network analysis.
The Link Probability Model (LPM) can be used as an alternative to
Exponential Random Graph Models (ERGM) to simulate network data. The
LPM characterizes the networks in terms of link probabilities based on
historical frequencies. In this paper, the LPM is presented, compared and
contrasted with the ERGM. The relative utility of the two approaches is
examined by applying both to four longitudinal data sets. The relative
strengths and weaknesses of the two approaches in terms of data requirements,
scalability, and assumptions are described.
21 pages
*School of Information Science, Curtin University of Australia, Perth, Western
Australia
**Department of Statistics, Oxford University, Oxford, England
