Importing & Exporting

Real network data rarely arrives as an nwcommands-native object — it’s a text file, a UCINET matrix, or something a collaborator built in Pajek. This tutorial covers moving data across that boundary in both directions.

Importing a plain edgelist

nwimport reads a range of common formats, and filename can be a URL — no local download step needed:

. nwimport "https://raw.githubusercontent.com/thomas-u-grund/nwcommands/master/data/edgelist_example.txt", type(edgelist)
------------------------------
Importing successful

. nwsummarize
--------------------------------------------------
   Network name:  edgelist_example
   Network id:  1
   Directed: true
   Valued: false
   Two-mode: false
   Nodes: 4
   Selfloop: false
   Arcs: 5
   Minimum value:  0
   Maximum value:  1
   Density:  .417
   Temporal: false

The imported network is named after the file by default (edgelist_example) — use name() to pick something else.

Exporting to UCINET and Pajek

nwexport writes one network at a time, in either UCINET (.dl) or Pajek (.net) format:

. nwwebuse florentine, nwclear

. nwexport flomarriage, type(ucinet) replace
Exporting network: flomarriage
Saved as file: flomarriage.dl

. nwexport flobusiness, type(pajek) replace
Exporting network: flobusiness
Saved as file: flobusiness.net

By default the file lands in Stata’s current working directory, named after the network; fname() overrides that.

Round-tripping through a plain edgelist

Sometimes the format you need isn’t a file format at all — you just want the network as an ordinary Stata dataset to reshape, filter, or merge with something else, then turn back into a network. nwtoedge does the first half:

. nwwebuse glasgow, nwclear

. nwtoedge glasgow1
(50 real changes made)

. describe

Contains data from /var/folders/hx/2zp2k5ps2y7d_6hx015fcr1h0000gn/T//S_50827.000001
 Observations:         2,500                  
    Variables:             3                  5 Sep 2026 14:21
------------------------------------------------------------------------------------------------------------------
Variable      Storage   Display    Value
    name         type    format    label      Variable label
------------------------------------------------------------------------------------------------------------------
_ego            str3    %9s                   
_alter          str3    %9s                   
glasgow1        byte    %10.0g                
------------------------------------------------------------------------------------------------------------------
Sorted by: _ego  _alter

. list in 1/5

     +--------------------------+
     | _ego   _alter   glasgow1 |
     |--------------------------|
  1. |   n1       n1          . |
  2. |   n1      n10          0 |
  3. |   n1      n11          1 |
  4. |   n1      n12          0 |
  5. |   n1      n13          0 |
     +--------------------------+

Note that this is every possible pair of the 50 nodes (2,500 rows = 50×50), not just the pairs that are actually tied — the tie value itself is the column named after the network (glasgow1, here 0/1/missing for the diagonal). Keep only the real ties and rebuild:

. keep if glasgow1 == 1
(2,387 observations deleted)

. nwfromedge _ego _alter, name(rebuilt)

. nwsummarize rebuilt
--------------------------------------------------
   Network name:  rebuilt
   Network id:  4
   Directed: true
   Valued: false
   Two-mode: false
   Nodes: 50
   Selfloop: false
   Arcs: 113
   Minimum value:  0
   Maximum value:  1
   Density:  .046
   Temporal: false

rebuilt comes back with all 50 of glasgow1’s original nodes, isolates included, even though an edgelist itself has no way to represent a zero-tie node as a row of its own — three of Glasgow’s 50 students have no friendship ties at all. nwtoedge attaches its source network’s own full node list to the edgelist dataset behind the scenes (no extra rows or variables to see), and nwfromedge automatically adds back anything on that list that the edgelist’s own ties did not reproduce - no manual bookkeeping needed. This only works because nwtoedge produced this particular edgelist; an edgelist typed in by hand or brought in from another source carries no such list, so genuinely use nwaddnodes for that case, same as always.