Diagnosing Multicollinearity in Exponential Random Graph Models

Abstract

Exponential random graph models (ERGM) have been widely applied in the social sciences in the past ten years. However, diagnostics for ERGM have lagged behind their use.Collinearity-type problems can emerge without detection when fitting ERGM, yielding inconsistent model estimates and problematizing inference from parameters. This article provides a method to detect multicollinearity in ERGM. It outlines the problem and provides a method to calculate the variance inflation factor from ERGM parameters. It then evaluates the method with a Monte Carlo simulation, fitting 216,000 ERGMs and calculating the variance inflation factors for each model. The distribution of variance inflation factors is analyzed using multilevel regression to determine what network characteristics lend themselves to collinearitytype problems. The relationship between variance inflation factors and unstable standard errors (a standard sign of collinearity) is also examined. The method is shown to effectively detect multicollinearity and guidelines for interpretation are discussed.

A matter of information – The influence of international bureaucracies in global climate governance networks

“…The goodness of fit diagnostics for both models were obtained according to Hunter, Goodreau, and Handcock (2008) and can be found in the Appendix. Moreover, we checked the model for multicollinearity by examining the variance inflation factors (Duxbury 2018) and found no issue with any parameter. As described above, our first model included actor types as a node level attribute, and on a dyadic level as homophily effect between actor types.…”

Emergent Relational Structures at a “Sharing Economy” Festival

“…Comparison of the network statistics measured in the original data and networks simulated from the fitted model [16] indicates good fit 5 . Multi-collinearity checks performed as per [12] do not point to major concerns.…”

Score-Driven Exponential Random Graphs: A New Class of Time-Varying Parameter Models for Dynamical Networks

“…That would necessitate a precise identification of areas of the parameter space that are free from inferential issues. The topic of multi-collinearity is recently receiving attention (see Duxbury, 2018), but a clear characterization for different values of the ERGM parameters is not yet available.…”