[Rsiena-help] Effects of covariate similarity on the evolution of a bipartite graph?

[Tom Snijders]

Dear Matteo,

That is an interesting topic. For the moment this effect has been blacklisted only because this distance-2 effect still needs to be checked once more (also in its one-mode version) and because the filtering of effects to allow them in the 2-mode version still is incomplete. You are right that it should be allowed in.
If you (or anyone) wishes to contribute and speed this up, then the first thing would be to check that the target statistics for this effect in its one-mode version as reported in the output are identical to what one should expect from the definition of the effect in the manual. This will be the first thing I would want to do as a first step to incorporate this effect in the two-mode version.
Best wishes,
Tom

================================================================
Tom A.B. Snijders
Professor of Statistics in the Social Sciences
Department of Politics and Department of Statistics
Nuffield College
University of Oxford
tel. +44-01865-278599


-----Original Message-----
From: rsiena-help-bounces at lists.r-forge.r-project.org [mailto:rsiena-help-bounces at lists.r-forge.r-project.org] On Behalf Of Matteo Gagliolo
Sent: 08 October 2012 15:33
To: rsiena-help at lists.r-forge.r-project.org
Subject: [Rsiena-help] Effects of covariate similarity on the evolution of a bipartite graph?

Dear RSiena users,

        I'm trying to analyze a two-mode network representing interlocking directorates. I have covariates for the mode corresponding to boards, and I would like to see if there is a tendency of directors to join boards with similar covariates (or equivalently, a tendency of boards to form interlocks with similar boards). Sort of an homophily at distance 2.

Taking the boards as mode one, I think this effect could be implemented as an interaction among egoX and altDist2, as the latter would represent the average X of boards on which a certain director is already sitting.

Here comes my problem: altDist2 is currently "blacklisted" in the code for bipartite nets. Diving into the code (RSienaTest r. 220) I found out where it all happens: in file effects.r:773, function covarBipartiteEff (comments are mine, to explain the issue):
---
covarBipartiteEff<- function(covarname, poszvar, moreThan2, nodesetNbr,
                                 name)
    {
        covRateEffects  <-  NULL
        if (nodesetNbr == 1)
        {#This is the branch taken as the covariate is on the 1st mode(boards)
            covObjEffects <-
                createEffects("covarBipartiteObjective", covarname,
                              name=varname,
                              groupName=groupName, group=group,
                              netType=netType) #This returns "altX"  "altSqX"
"egoX"    "altDist2" "simDist2", each replicated as eval/creat/endow
            covObjEffects <-
                covObjEffects[covObjEffects$shortName %in%
                              c("egoX"), ] #Here only "egoX" effects are selected
 ---

So, it seems the code is currently allowing only for egoX effects for covariates of the 1st mode.

Does anyone have an idea why? As all distance 2 nodes are in the same mode, I don't see why altDist2 and simDist2 should be blacklisted.

Or maybe there is a more correct way to implement the "distance 2 homophily" I am interested in?

Thanks,

Matteo

--
Matteo Gagliolo <mgagliol at ulb.ac.be>
Group for research on Ethnic Relations, Migration and Equality (GERME), Institute of Sociology; Machine Learning Group (MLG), Computer Science Department.
Université Libre de Bruxelles (ULB).
GERME, ULB, CP 124, 44 av. Jeanne
1050 Brussels, Belgium
Phone +32 2 650 4798, Fax +32 2 650 4659 http://homepages.ulb.ac.be/~mgagliol/

_______________________________________________
Rsiena-help mailing list
Rsiena-help at lists.r-forge.r-project.org
https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/rsiena-help

Nuffield College is a Registered Charity No. 1137506. Registered Office: Nuffield College, New Road, Oxford, OX1 1NF