nwergm

Exponential random graph models (ERGM) for Stata: why a network looks the way it does, and which local patterns (reciprocity, clustering, homophily) build it.

Install Start with the intro tutorial Manual nwergm on GitHub

What nwergm is

nwergm fits exponential-family random graph models to a network that is observed once. The model describes the probability of the whole network as a function of a few network statistics (the number of ties, of reciprocated pairs, of triangles, of ties between people of the same group, …), each with a coefficient that is estimated from the data. A coefficient tells you whether the pattern occurs more or less often than expected by chance, given the other patterns in the model. nwergm estimates the models by maximum pseudolikelihood where this is exact and by Monte Carlo maximum likelihood otherwise, and it provides the tools to check, compare and simulate from the fitted model. It is a port of the R package ergm of the statnet project to Stata and is part of the nwcommands family of packages.

Who it is for

nwergm is for people who work in Stata, or prefer it to R, and who want to fit ERGMs there. The tutorials start from a plain-language explanation, then run a real command and read the real output. It does not aim to replace ergm. R’s ergm is the reference implementation and the more complete program, and the page nwergm and R’s ergm says exactly what nwergm has and what it lacks.

Key facts

   
Terms 125 of the 139 terms of R’s ergm 4.12.0 (89.9%), and 125 of the 126 that apply to binary networks (99%); the missing ones are the 13 terms for valued networks and the operator B, see nwergm and R’s ergm.
Agreement with ergm for each term, the observed statistics, the maximum pseudolikelihood estimates and the Monte Carlo maximum likelihood estimates were compared with ergm 4.12.0
Estimators maximum pseudolikelihood (also penalized and other variants), Monte Carlo maximum likelihood, contrastive divergence, stochastic approximation
Networks directed, undirected and two-mode (bipartite), binary ties
Missing data missing dyads handled as in R (Handcock and Gile); missing covariate values are refused, as in R
Fit statistics log-likelihood, deviance, AIC, BIC, estat ic, lrtest, predict
Constraints sampling and observation constraints (constraints(), obsconstraints())
Dialog boxes db nwergm sets every term and option; db nwergm_simulate and db nwergm_estat cover simulation and postestimation (from version 1.1.1)
Postestimation goodness of fit (estat gof), MCMC diagnostics (estat mcmcdiag), simulation from a fit (nwergm simulate)
Engine a compiled plugin for the Monte Carlo sampler (macOS, Linux, Windows)
Speed Faster than R’s ergm on all eight of our benchmark models (2.9 to 9.4 times, medians of seven fits); the final Monte Carlo sample at the default target is smaller than ergm’s; see Speed compared with R
Version 1.3.5 (28 October 2026)
License GNU General Public License, version 3

The operator terms (operators()), the term balance, the curved two-mode terms and altkstar with an estimated decay were added in version 1.1.0; earlier releases do not have them.

In this section

The core nwcommands is introduced in the tutorials; the other model packages are described in Intro to Longitudinal Network Models.

Table of contents

Back to top

nwcommands is free software under the GNU General Public License, version 3, as are the model packages nwsaom, nwergm, nwdynam and nwrem. Citing and licenses · Get help · GitHub. Issues with the core go to the nwcommands repository; issues with a model package go to that package's own repository.

This site uses Just the Docs, a documentation theme for Jekyll.