nwsaom

Stochastic actor-oriented models (SAOM) for Stata: how networks and the behavior of the actors change over time.

Install Start with Tutorial 1 Manual nwsaom on GitHub

What nwsaom is

nwsaom fits stochastic actor-oriented models (Snijders 2001) to networks that were observed at two or more points in time, such as a school class surveyed every year. The model describes how the network changed between the waves as the result of the choices of the actors, each of whom creates, keeps or drops its own ties according to a few simple rules (they reciprocate, they close triangles, they prefer similar others, …), and it estimates how strong each rule is. The behavior of the actors (drinking, smoking, opinions) can change together with the network, which lets you separate selection (people choose friends who are like them) from influence (friends become alike). nwsaom is a port of the R package RSiena to Stata and is part of the nwcommands family of packages.

Who it is for

nwsaom is for people who work in Stata, or prefer it to R, and who are new to these models. The tutorials start from a plain-language explanation, then run a real command and read the real output. nwsaom follows the R package RSiena and is not a replacement for it: RSiena is the reference implementation and the more complete program, and the page nwsaom and RSiena says exactly what nwsaom has and what it lacks. nwsaom is not affiliated with or endorsed by the authors of RSiena; errors in nwsaom are the author’s.

Key facts

Effects	about 94% of the effects in the effects table of RSiena 1.6.6 that are planned for nwsaom (1,058 of 1,121 rows; effects added in RSiena 1.6.7 to 1.6.14 are implemented too but not counted, and the counts change with RSiena releases); the remaining rows of that table are accounted for in nwsaom and RSiena
Agreement with RSiena	every implemented effect was compared with siena07, by the observed statistic and by the estimates
Estimators	method of moments (conditional and unconditional), maximum likelihood, Bayesian
Networks	directed, undirected, two-mode, several networks at once, network-behavior co-evolution
Data features	covariates, composition change, missing data, structural zeros
Engine	a compiled, multithreaded plugin for macOS, Linux and Windows
Postestimation	goodness of fit, score tests, time-heterogeneity test, relative importance, simulation
Dialog boxes	db nwsaom (every option and effect), db nwsaom_multinet, db nwsaom_bayes, db nwsaom_meta, db nwsaom_describe, db nwsaom_estat (from version 1.0.3)
Speed	On one thread faster than RSiena in all 28 benchmark models of the article (median 2.1 times); see Speed compared with R
Version	1.3.4 (28 October 2026)
License	GNU General Public License, version 3

In this section

• Install: core first, then nwsaom; releases, updating, the plugin.
• Tutorial 1: A first SAOM: a first model for the Glasgow friendship data, and how to read the table.
• Tutorial 2: Convergence: what the convergence line means and what to do when a model does not converge.
• Tutorial 3: Goodness of fit: estat gof, and improving a poor model.
• Tutorial 4: Selection and influence: friendship and drinking together.
• Tutorial 5: Testing and comparing: estat score, estat timetest, estat importance, fix().
• Tutorial 6: Beyond the basic model: maximum likelihood, Bayesian estimation, several networks.
• Tutorial 7: Changing actor sets and missing data: present(), joinat()/leaveat(), missnet(), missbeh(): actors who join or leave, and missing data.
• Tutorial 8: Time-varying and dyadic covariates: tvcov() and dyadcov(): covariates that change over time and covariates of pairs.
• Tutorial 9: Undirected and two-mode networks: undirected networks (symtype()) and two-mode networks.
• Tutorial 10: Endowment, creation, interactions and rate effects: endow()/creation(), interact(), ratecov(), outrate.
• Tutorial 11: Simulation and what-if analysis: estat simulate, nwsaom simulate, estat mems.
• Tutorial 12: Conditional and unconditional estimation, threads and speed: conditional and unconditional estimation, cores(), k0()/k3().
• Tutorial 13: Combining several groups with a meta-analysis: nwsaom meta for several groups (RSiena’s siena08).
• Tutorial 14: Several groups at once: one model for all groups with groups() and gwaves() (RSiena’s sienaGroupCreate).
• nwsaom and RSiena: what is the same, what differs, a translation table.
• Effects: an overview of the effect families.
• How to cite

Links

• Source code and releases: github.com/thomas-u-grund/nwsaom
• The manual (PDF), archived at Zenodo: doi:10.5281/zenodo.23623807; also in the GitHub repository
• Stata’s own documentation: help nwsaom, help nwsaom_estat, help nwsaom_remarks; or the nwsaom reference page
• The paper on nwsaom is in preparation; the current version is on the literature page.

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

---

Table of contents

• Install
• Tutorial 1: A first SAOM
• Tutorial 2: Convergence
• Tutorial 3: Goodness of fit
• Tutorial 4: Selection and influence
• Tutorial 5: Testing and comparing
• Tutorial 6: Beyond the basic model
• Tutorial 7: Changing actor sets and missing data
• Tutorial 8: Time-varying and dyadic covariates
• Tutorial 9: Undirected and two-mode networks
• Tutorial 10: Endowment, creation, interactions and rate effects
• Tutorial 11: Simulation and what-if analysis
• Tutorial 12: Conditional and unconditional estimation, threads and speed
• Tutorial 13: Combining several groups with a meta-analysis
• Tutorial 14: Several groups at once
• nwsaom and RSiena
• Effects
• How to cite
• Manual