Activity Number:
|
695
|
Type:
|
Contributed
|
Date/Time:
|
Thursday, August 4, 2016 : 10:30 AM to 12:20 PM
|
Sponsor:
|
Section on Statistical Graphics
|
Abstract #318914
|
View Presentation
|
Title:
|
Improved Simulation for Exponential Random Graph Models for Social Network Analysis
|
Author(s):
|
Junchi Guo* and Michael Larsen
|
Companies:
|
and The George Washington University
|
Keywords:
|
ERGM ;
Stratified Sampling Proposal ;
MCMC convergence ;
Markov chain mixing ;
Multiple Chains
|
Abstract:
|
Exponential Random Graph Models (ERGMs) are one class of models used to describe observations from a social network. ERGMs have been used in many applications and are popular due to their ability to model dependence between connections in a network. Sufficient statistics for the models are called network statistics. Their computation relies on MCMC sampling. Guo and Larsen (2015) introduced stratified sampling proposals to make the Metropolis Hasting more efficient for ERGMs using the GWESP network statistic. This talk presents extensions on this topic involving variations and refinements of sampling proposals for MCMC sampling. The improved MCMC algorithms also increase the feasibility of using multiple chains with dispersed starting networks to monitor the mixing of the Markov chains used in sampling. Convergence monitoring methods for ERGMs are described and illustrated.
|
Authors who are presenting talks have a * after their name.