|
Activity Number:
|
430
|
|
Type:
|
Invited
|
|
Date/Time:
|
Thursday, August 7, 2003 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Bayesian Stat. Sciences
|
| Abstract - #300802 |
|
Title:
|
Mixed Effects p* Model for Multiple Social Networks
|
|
Author(s):
|
Jonathon L. Templin*+ and Moon-ho Ringo Ho and Carolyn J. Anderson and Stanley Wasserman
|
|
Companies:
|
University of Illinois and University of Illinois at Urbana-Champaign and University of Illinois and University of Illinois
|
|
Address:
|
603 E Daniel St., Champaign, IL, 61820-6232,
|
|
Keywords:
|
network analysis ; p* models ; MCMC for networks
|
|
Abstract:
|
The p* model for social networks was originally developed for analyzing a single network. Anderson et al. (1999) extended the model for multiple networks; however, their extension is only feasible for a small number of networks and does not distinguish between random and systematic differences between networks. We propose an extension of the p* model for modeling multiple networks where the networks are a random sample from a population of networks. Rather than treating model parameters as fixed, some or all of the parameters can be modeled as a function of other variables. The models are essentially random coefficient logistic regression models that can be fit by available software. The mixed effects p* model can be used to investigate similarities between networks while allowing for random differences. The models permit generalizations to the population of networks from which the sample of networks was drawn. A sensitivity analysis of both fixed and random effects parameters is performed by simulating random networks through use the MCMC algorithm by Snijders (2002). As an example, models are fit to friendship networks in 37 classes nested within 15 schools (Ryan, 2002).
|
