Installing nwsaom

nwsaom is a separate package (GPL-3) that needs the core nwcommands. Install the core first, then nwsaom.

1. The core

. net install nwcommands, from(https://nwcommands.org/net/) replace
. nwinstall, everything

nwinstall, everything installs the core and all four model packages in one step; nwinstall, all installs the core only.

See Installation & Setup for details, including computers without internet access or administrator rights.

2. nwsaom

. nwinstall, models

If you used nwinstall, everything above, the model packages are already installed; after nwinstall, all (core only) add them with nwinstall, models. Repeating either command updates them. To install nwsaom only, or a specific tagged version, use its own address:

. net install nwsaom, from(https://raw.githubusercontent.com/thomas-u-grund/nwsaom/main/)

(nwinstall, modelpkg(nwsaom) does the same from the core.)

A specific release

Every release is tagged on GitHub and each tag has its own address. For version 1.3.4:

. net install nwsaom, from(https://raw.githubusercontent.com/thomas-u-grund/nwsaom/v1.3.4/) replace

Use a tagged release when you want an analysis to be reproducible: you can then say exactly which version produced the results.

3. Check the installation

. nwsaom version
nwsaom version 1.3.4, 28oct2026
  core package nwcommands : installed (development copy, no distribution date)
  native plugin           : .../saom_sim_macos.plugin (plugin build 140)
  compared with           : R RSiena 1.6.14
  license                 : GNU GPL-3

The output shows the release of nwsaom, the core, the native plugin and the version of RSiena that the results were compared with. (The lines about the core show “development copy” here because the examples on this website were run from a source checkout; a normal installation shows the distribution date of the core.) The release is stored in r(version), and after a fit in e(version); quote it when you report an analysis. nwversion lists the core and all installed model packages. help nwsaom opens the complete documentation.

Updating

Repeat nwinstall, models (or nwinstall, everything), or, for this package only, its net install line with replace:

. net install nwsaom, from(https://raw.githubusercontent.com/thomas-u-grund/nwsaom/main/) replace

Updating the core never overwrites nwsaom, and updating nwsaom never changes the core. net describe nwsaom shows the installed version and distribution date.

The native plugin

The simulations at the heart of an SAOM run in a compiled plugin that nwsaom ships for macOS, Linux and Windows. It is installed together with the package and found automatically; there is nothing to configure, and it uses all physical cores of the computer by default (cores(1) runs it on a single thread; the results for a given seed() are the same whatever the number of threads). nwsaom version shows the path of the plugin in use.

If no plugin can be found, nwsaom version says “missing”, and the fit is run by a slower engine written in Mata wherever that is possible. Not every effect has a Mata implementation: some effects, and the estimators and tools that need the plugin (for example nwsaom multinet, estat importance and nwsaom bayes), stop with a message instead. After a fit, e(engine) says which engine ran (native or mata).

If you have an old copy of nwsaom in Stata’s PLUS folder as well as a new one elsewhere on the ado-path, Stata uses the first one it finds; if an update does not seem to take effect, which nwsaom and nwsaom version show which copy is in use.

Requirements

Stata 17 or later; Stata/BE is enough. Tested with Stata 17 and 18 (the Mata libraries are built with Stata 17, the oldest tested version; Stata 16 and older are untested). The plugin exchanges data with Mata through frames. The examples used Stata/BE 18.
The core nwcommands, as above.
No R and no other software.

Tutorial 1 fits a first model.