`nwfromedge`

Import network data from an edge list

Syntax

nwfromedge 
 fromid
 toid
[tievalue]
[if]
[,
name(newnetname)
xvars
labs(lab1 lab2 ...)
undirected
directed
twomode
valued
unvalued]


`name(newnetname)`	name of the new network; default = network
`xvars`	generate Stata variables for the network
`labs`(lab1 lab2 …)	overwrite node labels
`undirected`	force the network to be undirected (alias: `forceundirected`)
`directed`	force the network to be directed (alias: `forcedirected`)
`twomode`	declare a two-mode (bipartite) network instead - fromid/toid are the mode-1/mode-2 id variables, not a directed ego/alter pair. An exact alias for nw2fromedge, forwarding `name()`/`xvars` only; cannot be combined with `directed`/`undirected`/`forcedirected`/`forceundirected`. See nw2fromedge for the full two-mode-specific behavior (same-label disambiguation, mode assignment, `project()`)
`valued`	mark the network as valued (default if a value variable is given, also when every value is 1)
`unvalued`	mark the network as binary even if a value variable is given
`noclear`	do not clear existing dataset
`replace`	if a network named newnetname already exists, drop it and use this name anyway (see nwset for the same convention)
`labprefix`(string)	prefix used for auto-generated node labels when `fromid`/`toid` are numeric and `labs()` is not specified; default = n - named `labprefix()`, not `prefix()`, to avoid colliding with nwrecode’s unrelated `prefix()` (which prefixes new network names, not node labels)
`overwrite`	forwarded to nwload governing whether this command’s own generated Stata variables overwrite existing ones of the same name - unrelated to `replace` above, which is about the network, not Stata variables

Description

nwfromedge imports a network from a dataset in edgelist format.

An edgelist or arclist is a set of two (or three in the case of a valued network) variables representing relations. Nodes are identified by entries in the cells. For example, the data

   +---------------+
   |  fromid  toid |
   |---------------|
1. |  1       2    |
2. |  2       3    |
3. |  4       2    |
   +---------------+

stores information about three ties (1=>2), (2=>3) and (4=>2) among four unique network nodes. The variables defining the edges can also be string variables.

   +--------------------------+
   |  fromid    toid     value|
   |--------------------------|
1. |  Peter     Thomas   1    |
2. |  Tim       Peter    3    |
3. |  Mathilde  Thomas   2    |
   +--------------------------+

Here, there are also three relationships: (Peter => Thomas), (Tim => Peter) and (Mathilde => Thomas).

The following command declares such data as network data:

. nwfromedge fromid toid value, name(mynet)

This automatically generates the relevant meta-information for the network and makes it available for other programs under the netname mynet. In case no name() is specified, the command tries to come up with a suitable name for the new network. By default, it tries network, however, if a network with this name already exists, it comes up with an alternative name network_1 and so on (see nwvalidate).

After a network has been declared, one can refer to it by its netname, just as if one would refer to a varname. For example, this makes a network plot of mynet.

. nwplot mynet

Or alternatively, this calculates the betweenness centrality of the nodes in mynet.

. nwbetween mynet

By default, nwfromedge recognizes if a network is directed or undirected, i.e. for each dyad entry (i,j) there is also a dyad entry (j,i). However, this automatic detection can be overwritten with the options undirected and directed. With undirected, a tie that is listed in only one direction is stored as it is (a negative value stays negative); if both directions are listed with different values, the larger value is kept.

One can also transfrom any network that exists in memory into such an edgelist with nwtoedge.

Examples

This loads a network dataset from the internet and transforms the network glasgow1 into an edgelist.

. nwwebuse glasgow, nwclear
. nwtoedge glasgow1

nwtoedge produces variables _ego/_alter (not _fromid/_toid) plus a tie-value column named after the network itself (glasgow1 here, not _link) - and, being every possible pair rather than a compact edgelist, it needs filtering down to the real ties before being turned back into a network:

. keep if glasgow1 == 1
. nwfromedge _ego _alter, name(mynet)

glasgow1 has 50 nodes but only 47 with at least one tie; mynet still comes out with all 50 - the 3 isolates nwtoedge could not write as edgelist rows are added back automatically, with no extra step needed (see this file’s own Supported network types section above). This only works because nwtoedge itself produced this particular edgelist; it would not apply to an edgelist typed in by hand or imported from an external source.

Supported network types

Binary: yes. Directed: yes, via directed/undirected/forcedirected/forceundirected. Weighted: yes - a third edge-list column supplies tie values. Signed: yes - with undirected a tie listed once keeps its sign. Two-mode: yes - with twomode, an exact alias for nw2fromedge, which builds the network from the two id columns. A node with zero ties (an isolate) can never appear as a ROW in an edgelist - there is no pair to write - but a round trip through nwtoedge specifically is not affected: nwtoedge attaches its source network’s own full node list to the edgelist dataset (invisibly - no extra variables or rows), and nwfromedge automatically adds back any node that list has but the edgelist’s own ties did not reproduce. A hand-built or externally-sourced edgelist carries no such list, so isolates neither of those every had are genuinely unrepresentable - add them with nwaddnodes same as always.

nwcommands version 3.17.6, 9 November 2026 (check your installed version)