|
Activity Number:
|
284
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Tuesday, August 4, 2009 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Survey Research Methods
|
| Abstract - #305149 |
|
Title:
|
Towards a Unified Framework for Inference with Aggregated Relational Data
|
|
Author(s):
|
Tian Zheng*+ and Tyler H. McCormick
|
|
Companies:
|
Columbia University and Columbia University
|
|
Address:
|
Room 1007, New York, NY, 10027,
|
|
Keywords:
|
Aggregated relational data ; Latent space models ; Social networks
|
|
Abstract:
|
Aggregated Relational Data (ARD), originally introduced by Killworth et al. (1998) as "How many X's do you know" survey questions, are a common tool for observing social networks indirectly. Previous methods for ARD estimate specif c network features, such as overdispersion. We suggest a more general approach to understanding social structure using ARD based on a latent space framework. We first show ARD contain information about latent structure by apply a primitive latent-space model to data from McCarty et al. (2001). This example also demonstrates the utility of these models for understanding the networks of individuals who are dicult to reach with traditional surveys. We then suggest using latent space models as a uni ed framework for inference with ARD by demonstrating that the network features estimated using previous methods can be represented as latent structure.
|
