|
Activity Number:
|
219
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, August 3, 2009 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Statistical Computing
|
| Abstract - #305431 |
|
Title:
|
Random Graph Models for Inference: The Devil Is in the Details
|
|
Author(s):
|
Elizabeth A. Beer*+ and Carey E. Priebe and Edward R. Scheinerman
|
|
Companies:
|
Johns Hopkins University and Johns Hopkins University and Johns Hopkins University
|
|
Address:
|
Dept. of App. Math & Stat, 302 Whitehead Hall, Baltimore, MD, 21218,
|
|
Keywords:
|
random graphs ; inference ; network theory
|
|
Abstract:
|
Statistical network theory depends on data in graph form: social networks, communication networks, and so on. We present a family of latent position random graph models designed for inference on such graph data (in particular, for detecting graphs with localized regions of anomalously high activity). Seemingly minor differences in the latent position spaces underlying these models have dramatic effects on limiting properties of the resulting random graphs; this in turn affects our ability to use such models for inference. We use simple statistics, including edge count, to illustrate the effects of small model changes.
|
