The Problem of Scaling in Exponential Random Graph Models

被引:30
作者
Duxbury, Scott W. [1 ]
机构
[1] Univ N Carolina, Sociol, Chapel Hill, NC 27515 USA
关键词
social network analysis; exponential random graph models; scaling; mediation; moderation; SOCIAL NETWORKS; UNOBSERVED HETEROGENEITY; LOGIT; PROBIT; FAMILY; COEFFICIENTS; REGRESSION; LIMITS; END;
D O I
10.1177/0049124120986178
中图分类号
O1 [数学]; C [社会科学总论];
学科分类号
03 ; 0303 ; 0701 ; 070101 ;
摘要
This study shows that residual variation can cause problems related to scaling in exponential random graph models (ERGM). Residual variation is likely to exist when there are unmeasured variables in a model-even those uncorrelated with other predictors-or when the logistic form of the model is inappropriate. As a consequence, coefficients cannot be interpreted as effect sizes or compared between models and homophily coefficients, as well as other interaction coefficients, cannot be interpreted as substantive effects in most ERGM applications. We conduct a series of simulations considering the substantive impact of these issues, revealing that realistic levels of residual variation can have large consequences for ERGM inference. A flexible methodological framework is introduced to overcome these problems. Formal tests of mediation and moderation are also proposed. These methods are applied to revisit the relationship between selective mixing and triadic closure in a large AddHealth school friendship network. Extensions to other classes of statistical work models are discussed.
引用
收藏
页码:764 / 802
页数:39
相关论文
共 59 条
[41]   Logistic Regression: Why We Cannot Do What We Think We Can Do, and What We Can Do About It [J].
Mood, Carina .
EUROPEAN SOCIOLOGICAL REVIEW, 2010, 26 (01) :67-82
[42]  
Newman Mark, 2018, Networks: An Introduction, DOI DOI 10.1093/ACPROF:OSO/9780199206650.001.0001
[43]   Connected in Crime: The Enduring Effect of Neighborhood Networks on the Spatial Patterning of Violence [J].
Papachristos, Andrew V. ;
Bastomski, Sara .
AMERICAN JOURNAL OF SOCIOLOGY, 2018, 124 (02) :517-568
[44]   The Corner and the Crew: The Influence of Geography and Social Networks on Gang Violence [J].
Papachristos, Andrew V. ;
Hureau, David M. ;
Braga, Anthony A. .
AMERICAN SOCIOLOGICAL REVIEW, 2013, 78 (03) :417-447
[45]   Consistent structure estimation of exponential-family random graph models with block structure [J].
Schweinberger, Michael .
BERNOULLI, 2020, 26 (02) :1205-1233
[46]   Local dependence in random graph models: characterization, properties and statistical inference [J].
Schweinberger, Michael ;
Handcock, Mark S. .
JOURNAL OF THE ROYAL STATISTICAL SOCIETY SERIES B-STATISTICAL METHODOLOGY, 2015, 77 (03) :647-676
[47]  
Snijders TA., 2002, J. Soc. Struct, V3, P1
[48]   The statistical evaluation of social network dynamics [J].
Snijders, TAB .
SOCIOLOGICAL METHODOLOGY 2001, VOL 31, 2001, 31 :361-395
[49]   New specifications for exponential random graph models [J].
Snijders, Tom A. B. ;
Pattison, Philippa E. ;
Robins, Garry L. ;
Handcock, Mark S. .
SOCIOLOGICAL METHODOLOGY 2006, VOL 36, 2006, 36 :99-153
[50]   DIRECT AND INDIRECT EFFECTS IN LINEAR STRUCTURAL EQUATION MODELS [J].
SOBEL, ME .
SOCIOLOGICAL METHODS & RESEARCH, 1987, 16 (01) :155-176