Core Concepts

Ordinary Stata holds one dataset in memory at a time. nwcommands adds a second kind of object alongside it — a network — and, unlike a dataset, you can have several networks loaded at once. This tutorial covers the model that makes that work: netnames, the current network, and how networks and your ordinary Stata variables share the same session without getting in each other’s way.

Multiple networks, one current network

Loading a network doesn’t clear anything — some datasets even declare more than one network at once. The Florentine families data (a classic small-network teaching example) ships with two: a marriage-tie network and a business-tie network.

. nwwebuse florentine

. nwset
(2 networks)
--------------------
      flobusiness
      flomarriage

Every command that takes a network defaults to whichever one you touched most recently — the current network. nwcurrent reports it, and switches it when given a name:

. nwcurrent
----------------------------------------
   Current network:  flomarriage
   Number of nodes:  16
----------------------------------------

. nwcurrent flobusiness
----------------------------------------
   Current network:  flobusiness
   Number of nodes:  16
----------------------------------------

. nwcurrent
----------------------------------------
   Current network:  flobusiness
   Number of nodes:  16
----------------------------------------

That last call takes no argument — it’s just asking “which one is current now” — and correctly still reports flobusiness, since nothing has changed it since the previous call.

netlist: referring to several networks at once

A netname is just what a network is called (flomarriage, glasgow1, whatever you name it). A netlist extends that to patterns — useful once a dataset has more than a couple of networks. glasgow is a three-wave friendship network among the same 50 pupils, loaded as three separate networks:

. nwwebuse glasgow, nwclear

. nwset
(3 networks)
--------------------
      glasgow1
      glasgow2
      glasgow3

The following four calls are all equivalent — an explicit list, a first-last range, a * wildcard, and the _all keyword:

. nwsummarize glasgow1 glasgow2 glasgow3
--------------------------------------------------
   Network name:  glasgow1
   Network id:  1
   Directed: true
   Valued: false
   Two-mode: false
   Nodes: 50
   Selfloop: false
   Arcs: 113
   Minimum value:  0
   Maximum value:  1
   Density:  .046
   Temporal: false
--------------------------------------------------
   Network name:  glasgow2
   Network id:  2
   Directed: true
   Valued: false
   Two-mode: false
   Nodes: 50
   Selfloop: false
   Arcs: 116
   Minimum value:  0
   Maximum value:  1
   Density:  .047
   Temporal: false
--------------------------------------------------
   Network name:  glasgow3
   Network id:  3
   Directed: true
   Valued: false
   Two-mode: false
   Nodes: 50
   Selfloop: false
   Arcs: 122
   Minimum value:  0
   Maximum value:  1
   Density:  .05
   Temporal: false

. nwsummarize glasgow1-glasgow3

. nwsummarize glasg*

. nwsummarize _all

Each of those three produces the exact same three-network listing shown above — see netlist for the full pattern syntax (*, ?, ~, ranges).

Networks coexist with ordinary Stata variables

A loaded network is a separate object — it doesn’t touch your dataset’s variables at all, and your dataset’s variables don’t touch it. You can gen a new variable while a network is loaded, and it just sits there as an ordinary column, right alongside whatever node-level variables the dataset already had:

. nwclear

. nwwebuse florentine

. gen note = "wealthiest families"

. describe

Contains data from https://raw.githubusercontent.com/thomas-u-grund/nwcommands/master/data/florentine.nwdta
 Observations:            16                  
    Variables:             6                  16 May 2019 17:02
------------------------------------------------------------------------------------------------------------------
Variable      Storage   Display    Value
    name         type    format    label      Variable label
------------------------------------------------------------------------------------------------------------------
_nwnode         str12   %12s                  
_nwinclude      float   %9.0g                 
wealth          float   %9.0g                 
priorates       float   %9.0g                 
seat            float   %9.0g                 
note            str19   %19s                  
------------------------------------------------------------------------------------------------------------------
Sorted by: 
     Note: Dataset has changed since last saved.

. list note in 1

     +---------------------+
     |                note |
     |---------------------|
  1. | wealthiest families |
     +---------------------+

_nwnode (the node label for each row) and _nwinclude came with the dataset; wealth, priorates, and seat are real node-level attributes from the original study; note is the one we just added. The network itself (flomarriage/flobusiness, still loaded) isn’t represented as columns at all — it lives as its own object, which is exactly why building or analyzing a large network never has to spend Stata’s own variable budget unless you deliberately load it as variables (see nwload).