`nwcontext`

Create a context variable

Syntax

nwcontext 
[netname]
,
attribute(varname)
[stat(statistic)
mode(context)
generate(newvarname)
mat(string)
noweight
replace
force]


`attribute(varname)`	Attribute variable
`stat`(statistic)	Statistic that is used to calculate context variable for node i from attributes of network neighbors
`mode`(context)	Define network neighbors of node i as either nodes j who receive ties from i, send ties to j or both
`generate(newvarname)`	Required. Name of the context variable to be generated
`mat(string)`	Store the result in a Mata matrix of this name instead of generating a Stata variable
`noweight`	Ignore valued ties and treats all as binary
`replace`	Replace `generate()`’s target variable if it already exists
`force`	Run the command although its matrices would not fit in memory (it stops with r(484) above 12 GB otherwise)

Description

nwcontext generates a context variable, which holds information about the attribute values of a node’s network neighbors.

For each node the set of network neighbors is specified in mode(), by default context = outgoing. This means that the new context variable newvarname[i] is calculated based on an all values varname[j], for which there is a tie from node i to node j. A network neighborhood of node i can be either 1) all nodes j to whom i has outgoing ties (default), 2) all nodes j from whom i receives incoming ties, or 3) both. In the case of valued ties varname[j] is weighted accordingly.

After that a statistic defined in stat() is calculated from all weigthed varname[j]. By default, the mean is calculated. More formally:

newvarname[i] = stat(varname[j]), for all j with y_ij > 0

When used with the defaults, the command calculates for each node i the mean score of its network neighbors on varname. Valid statistics are: the** mean, the **max, the min, the sum and the sd of varname.

Sometimes, however, one might want to calculate statistics including the attribute of ego. One can achieve this by using the ego-extended version of each statistic (meanego, maxego, minego, sumego and sdego), which are calculated in this way:

newvarname[i] = stat(varname[j]), for all j with y_ij > 0 or j == i

Remarks

In the case of undirected networks, no * mode* option needs to be specified.

By default, nodes with missing attributes are excluded from the calculation.

Examples

This example loads the Florentine marriage data. The variable wealth indicates how rich each family is. ** nwcontext** generates different variables: w_avg = average wealth of network neighbors, w_min = wealth of poorest network neighbor, w_min = wealth of richest network neighbor, w_sd = standard deviation of wealth over network neighbors.

. nwwebuse florentine
. nwcontext flomarriage, attribute(wealth) generate(w_avg)
. nwcontext flomarriage, attribute(wealth) generate(w_min) stat(min)
. nwcontext flomarriage, attribute(wealth) generate(w_max) stat(max)
. nwcontext flomarriage, attribute(wealth) generate(w_sd) stat(sd)

. list w*

    +----------------------------------------------------+
    | wealth      w_avg      w_min      w_max       w_sd |
    |----------------------------------------------------|
 1. |     10        103        103        103          0 |
 2. |     36   47.66667          8        103   40.33471 |
 3. |     55       61.5         20        103       41.5 |
 4. |     44   67.66666          8        146   57.86382 |
 5. |     20   83.33334         49        146   44.37968 |
    |----------------------------------------------------|
 6. |     32         36         36         36          0 |
 7. |      8       42.5         36         48   4.330127 |
 8. |     42          8          8          8          0 |
 9. |    103         31         10         55   17.26268 |
10. |     48         10         10         10          0 |
    |----------------------------------------------------|
11. |     49         70         20        146     54.626 |
12. |      3          .          .          .          . |
13. |     27         99         48        146   40.10819 |
14. |     10       75.5         48        103       27.5 |
15. |    146         35         20         49   11.89538 |
    |----------------------------------------------------|
16. |     48         46          8        103   41.04469 |
    +----------------------------------------------------+

You can plot the florentine marriage network with wealth as node label to better understand how these values come about:

. nwplot flomarriage, label(wealth)

Supported network types

Binary: yes. Directed: yes, via mode(outgoing|incoming|both|either). Weighted: yes by default (neighbor attribute values weighted by tie strength); noweight treats all ties as binary. Signed: not checked. Two-mode: same as for the bipartite graph read as a one-mode network (valid; no mode-specific normalisation).

nwcontext