Print

Print


*****  To join INSNA, visit http://www.insna.org  *****

Hi Kamal,

The answer depend on what kind of terms interactions of which you are
interested in. Essentially, in ERGMs you might have

1. "exogeneous" terms, e.g. edges, effects of node or edge attributes, etc.
2. "endogeneous" terms, e.g. terms to model transitivity,
(geometrically-weighted) degree distribution, etc.

If you are interested in introducing an interaction between terms that
both belong to group (1) it usually boils down to adding to the model
a (set of) properly constructed node/edge covariates. Creating
interaction effects between terms from group (1) and (2) or terms from
within (2) is not so trivial.

In R/statnet (1) is definitely possible with node[o/i]factor or
node[o/i]cov terms. In fact, it seems to me that what you wrote is an
interaction effect of two nodal attributes. These are dyad-independent
terms. If attributes are continuous, in statnet you could estimate
"null model" with just main effects (via node[i/o]cov terms). For the
second model compute a new node covariate as a product of these two
attributes, and add it as a 'node[i/o]cov' too. Do note that it is
something different than computing an edge attribute based on some
node attribute to model attribute-based proximity or social distance,
which in fact technically is also an interaction effect...

Above would be also possible in PNet as it is just a matter of
preparing node/edge covariates properly. No extra functionality
needed.

To model interaction effects involving endogeneous terms (2) in
statnet would require coding your own term which is possible (for
Windows https://statnet.org/trac/wiki/windowspackage), but less
trivial and more technical. I am not sure how far you can go with PNet
in that respect.


hth,
Michal



On Mon, May 9, 2016 at 3:23 PM, Kamal Badar <[log in to unmask]> wrote:
> ***** To join INSNA, visit http://www.insna.org *****
> Dear All,
>
> I have been wondering lately whether it is possible to test the
> moderating/interaction effect in ERGMs?
>
> It is understood that an attribute of a node might affect the probability of
> forming a new link. But can it be tested whether another attribute
> strengthens or weakens the aforementioned relationship? More specifically
> can we put the interaction term in ERGM models? If yes how do we interpret
> and plot the interaction effect?
>
> Any packages in R which do so? Any literature which tests the phenomenon?
>
> Help from the list will be highly appreciated.
>
> --
>
> Dr. Kamal Badar
> Assistant Professor (TTS)
> HEC Approved Ph.D. Supervisor
> Institute of Management Sciences
> University of Balochistan
> Quetta, Pakistan.
> http://sc.hec.gov.pk/aphds/Submit.asp?supid=6913
>
> _____________________________________________________________________ SOCNET
> is a service of INSNA, the professional association for social network
> researchers (http://www.insna.org). To unsubscribe, send an email message to
> [log in to unmask] containing the line UNSUBSCRIBE SOCNET in the body of
> the message.

_____________________________________________________________________
SOCNET is a service of INSNA, the professional association for social
network researchers (http://www.insna.org). To unsubscribe, send
an email message to [log in to unmask] containing the line
UNSUBSCRIBE SOCNET in the body of the message.