Exercise
SAOM
AdvancedSocialNetworkAnalysis
Background
Inthisexercise,IwantyoutousesomeofthedataIpresentedinthefirstsession.Thedataisabouta
friendshipnetworkinaDutchschoolclass.ThedatawerecollectedbetweenSeptember2003andJune
2004.
There are 26 students who were followed over their first year at secondary school during which
friendshipnetworksaswellasotherdatawereassessedatfourtimepointsatintervalsofthreemonths.
Therewere17girlsand9boysintheclass,aged11-13atthebeginningoftheschoolyear.Networkdata
were assessed by asking students to indicate up to twelve classmates which they considered good
friends.
Ultimately, I want you to run a network dynamics (only) model that replicates Model 0 (Table 1) in
Snijdersetal.(2010).Thequestioniswhatdrivesnetworkchangeinthisschoolclass?
Moreinformation
http://www.stats.ox.ac.uk/~snijders/siena/tutorial2010_data.htm
Data
http://www.stats.ox.ac.uk/~snijders/siena/klas12b.zip
YoucanfindthedatainthedatafolderthatIpreparedforthisworkshop.
Exercise1
-
-
-
-
-
-
-
Load the network datasets klas12b-net-1.dat, klas12b-net-2.dat, klas12bnet-3.dat and klas12b-net-4.dat fromthedatafolder.
read.table(...)isagoodstart
Transformthedata.frameobjectsyoujustgeneratedinmatrixobjectsandsavethesematrices
as net1, net2, net3 and net4.
Usethecommandas.matrix(...)
Thenetworksarecodedasfollows:Friendship:0=no,1=yes,9=missing,10=notamember
of the classroom (structural zero). In RSiena network values can only be 0,1,10,11 (structural
ones).
Therefore,youneedtodealwiththemissingcode9.Here,itmakessensetosetitto“structural
zero”.
a) Youcansubsetanetworkwiththissyntax:net1[net1==9]
b) Youcanassignwhateverisontherightsidetowhateverisontheleftside
oranarrow:a <- 10
Therefore,youneedtodealwiththemissingcode9.Here,itmakessensetosetitto“structural
zero”.
Next, load some additional data saved in klas12b-demographics.The fileklas12bdemographics.datcontainsthefollowingfourvariablessavedinfourcolumns.
§ Sex(1=girl,2=boy)
§ Age(years)
§ Ethnicity(1=Dutch,2=other,0=missing)
§ Religion(1=Christian,2=non-religious,3=non-Christianreligion,0=missing)
Generateanewobjectsex.dwhichholdsthefirstcolumnofthedata.frameyoujustloaded.
Youcanselectthefirstcolumnofadata.framecalleddatawithdata[,1]
Lastly,loadthedyadiccovariatesavedinklas12-b-primary.dat.Thisfilecontains
variablesindicatingiftwostudentswenttothesameprimaryschool.Savethisdataasamatrix
andcallitprimary.d § Sameprimaryschool:0=no,1=yes.
Great,ifyoumadeitsofarandhavesometimelefttrytoplotthenetworksrepresentedbythe
matricesnet1, net2, net3, net4.
Usegplot(...)toplotanetworkinmatrixformat
Exercise2
-
-
-
-
-
BeforeyoucanrunaRSienaanalyses,youneedtocreateasienaDependentobjectwhereyou
declarethesequenceofnetworks(net1… net4)asdependentvariable.Callthisnewobject
friendship
Checkthecodeweusedinthelabsessionforthesyntaxyouneedtouse.
Ifyoudonotknowthedimensionsofthenetworksyoucanfindthatoutwithe.g.
dim(net1)
Wewanttousesex.dasindependentvariable.Createaconstantcovariate(coCovar)object
forsex.dandcallitsex
Look-upthecodeweusedinthelabsessiontodosimilarthings.
Wealsowanttouseprimary.dasindependentvariable.Remember,primary.d isa
matrix(youshouldhavetransformeditintothisformat).Createaconstantdyadiccovariate
(coDyadCovar)objectforprimary.dandcallitprimary
Thesyntaxtodothisisverysimilartotheonewhenyoucreateaconstantcovariate.
Great,nowwehavealldependentandindependentvariableobjectsthatweneed.Youjust
needtoputitalltogetherandcreateasienaDataobjectandcallityourData.
Again,checkoutthecodeweusedinthelab.Youjustneedtoaddalldependentand
independentvariablestotheargumentslist.
ThecommandyouneedtouseissienaDataCreate
Finally,createasienaEffectsobjetfromthissienaDataobjectandcallityourEff
ThesimplecommandisgetEffects(…)
Exercise3
-
-
-
-
-
-
Wearenearlytheretorunamodel.Everythingisinplace.
Now, the only thing left to do is to specify the model you want to estimate. You do this by
altering the yourEff object you just created. The best way to include an effect is using
includeEffects(…)
Okay,wewanttoreplicatemodel0intable1ofSnijdersetal.(2010)
Thereareabunchofnetworkeffectsyouneedtoinclude.Theshortnamesfortheseeffectsare:
recip, transTrip, transTies, cycle3, inPopSqrt, outPopSqrt,
outActSqrt.
Youcanincludealloftheseeffectsinonego.Seehowwedidthisinthelabsession.
Next,includeegoX, altX, and sameX effectsfor sex.Remember,egoXjustmodels
thatindividualswithahighscoreonXsendmoretiesandaltXmodelsthatindivdiualswitha
highscoreonXreceivemoreties.AndsameXmodelsthattiesbetweentwoindividualswho
havethesamevalueonXaremorelikelytoform/maintainties.
Weusedtheseeffectsalreadyinthelabsession.Basically,youjustneedtoinclude
themintheargslistofinlcudeEffects(…)andspecifyinteraction1 =
“sex”
Lastly,letusincludeasimpledyadiccovariateeffectofprimary.Withthiseffectwewantto
modelthattiesmightbemorelikelytoform/maintainedwhentwoindividualsalreadywentto
thesameprimaryschooltogether.Wedidnotusesuchaneffectinthelabsessionbefore.This
istherelevantsyntax.Thegeneralformisalwaysthesame.Youjustneedtoknowtheshort
nameofaneffect(here:X)andspecifyitfurtherwithinteraction1(andsometimesalso
interaction2).
yourEff <- includeEffects(yourEff, X, interaction1 = "primary")
Yeah!!Andnow…runthemodel
yourAlgorithm <- sienaAlgorithmCreate(projname ='Exercise')
yourAns <- siena07(yourAlgorithm, data = yourData, effects
= yourEff)
LookatyourresultsyourAnsandbeexcitedthatyoujustreplicatedSnijdersetal.(2010)!
Yeah