We introduce a new method for estimating the parameters of exponential random
graph models. The method is based on a large-deviations approximation to the
normalizing constant shown to be consistent using theory developed by
Chatterjee and Varadhan. The theory explains a host of difficulties encountered
by applied workers: many distinct models have essentially the same MLE,
rendering the problems "practically" ill-posed. We give the first rigorous
proofs of "degeneracy" observed in these models. Here, almost all graphs have
essentially no edges or are essentially complete. We supplement recent work of
Bhamidi, Bresler and Sly showing that for many models, the extra sufficient
statistics are useless: most realizations look like the results of a simple
Erdos-Renyi model. We also find classes of models where the limiting graphs
differ from Erdos-Renyi graphs and begin to make the link to models where the
natural parameters alternate in sign.
Description
[1102.2650] Estimating and Understanding Exponential Random Graph Models
%0 Generic
%1 chatterjee2011estimating
%A Chatterjee, Sourav
%A Diaconis, Persi
%D 2011
%K exponential_model random_graphs statistics
%T Estimating and Understanding Exponential Random Graph Models
%U http://arxiv.org/abs/1102.2650
%X We introduce a new method for estimating the parameters of exponential random
graph models. The method is based on a large-deviations approximation to the
normalizing constant shown to be consistent using theory developed by
Chatterjee and Varadhan. The theory explains a host of difficulties encountered
by applied workers: many distinct models have essentially the same MLE,
rendering the problems "practically" ill-posed. We give the first rigorous
proofs of "degeneracy" observed in these models. Here, almost all graphs have
essentially no edges or are essentially complete. We supplement recent work of
Bhamidi, Bresler and Sly showing that for many models, the extra sufficient
statistics are useless: most realizations look like the results of a simple
Erdos-Renyi model. We also find classes of models where the limiting graphs
differ from Erdos-Renyi graphs and begin to make the link to models where the
natural parameters alternate in sign.
@misc{chatterjee2011estimating,
abstract = { We introduce a new method for estimating the parameters of exponential random
graph models. The method is based on a large-deviations approximation to the
normalizing constant shown to be consistent using theory developed by
Chatterjee and Varadhan. The theory explains a host of difficulties encountered
by applied workers: many distinct models have essentially the same MLE,
rendering the problems "practically" ill-posed. We give the first rigorous
proofs of "degeneracy" observed in these models. Here, almost all graphs have
essentially no edges or are essentially complete. We supplement recent work of
Bhamidi, Bresler and Sly showing that for many models, the extra sufficient
statistics are useless: most realizations look like the results of a simple
Erdos-Renyi model. We also find classes of models where the limiting graphs
differ from Erdos-Renyi graphs and begin to make the link to models where the
natural parameters alternate in sign.
},
added-at = {2011-02-21T07:54:21.000+0100},
author = {Chatterjee, Sourav and Diaconis, Persi},
biburl = {https://www.bibsonomy.org/bibtex/209b030b86e0f95987762c5bdd865b367/peter.ralph},
description = {[1102.2650] Estimating and Understanding Exponential Random Graph Models},
interhash = {697a37481901a107ac74dda68572bba3},
intrahash = {09b030b86e0f95987762c5bdd865b367},
keywords = {exponential_model random_graphs statistics},
note = {cite arxiv:1102.2650
Comment: 41 pages, 8 figures},
timestamp = {2011-02-21T07:54:21.000+0100},
title = {Estimating and Understanding Exponential Random Graph Models},
url = {http://arxiv.org/abs/1102.2650},
year = 2011
}