Activity Number:
|
45
- Emerging Methods for Network Testing and Related Problems
|
Type:
|
Invited
|
Date/Time:
|
Sunday, July 28, 2019 : 4:00 PM to 5:50 PM
|
Sponsor:
|
IMS
|
Abstract #300108
|
Presentation
|
Title:
|
Goodness-of-Fit Tests for 3 Variants of the Stochastic Block Model
|
Author(s):
|
Vishesh Karwa and Debdeep Pati and Sonja Petrovic* and Liam Solus and Mateja Raic and Dane Wilburne and Nikita Alexeev and Robert Williams and Bowei Yan
|
Companies:
|
Temple University and Texas A&M University and Illinois Institute of Technology and KTH, Sweden and University of Illinois at Chicago and ICERM, Brown University and unknown and Texas A&M University and University of Texas
|
Keywords:
|
testing model fit;
latent variables;
stochastic blockmodel
|
Abstract:
|
Stochastic block models (SBMs) with unknown block structure are widely used in analysis of real-world network data. Testing goodness of fit of such models is an important practical question. We develop finite-sample goodness-of-fit tests for three different variants of SBMs with unknown block assignments. The main building block for the goodness-of-fit test is an exact test for SBM with observed block assignment, which is implemented using tools from algebraic statistics. The methodology extends to any mixture of log-linear models on discrete data.
The talk will discuss the basics of the test setup, formal proofs of validity, and showcase some test power simulations.
|
Authors who are presenting talks have a * after their name.