[Rsiena-help] multiple networks

[Ruth M. Ripley]

Dear Maarten,

I suspected this might be difficult to explain, so did not try to write a
quick email on Wednesday!

I'll note replies below: [[rr]]

On Fri, 23 Apr 2010, Maarten Van Zalk wrote:

> Dear Ruth,
>
> To clarify your answer, I'd just like to ask about some of the terms
> you use (I'm not familiar with some of them):
>
> What is the difference (if any) between bipartite networks and
> multiple networks? Is it that bipartite networks you have the same
> egos and alters but there are different types of relationships between
> them, whereas with multiple networks the alters differ between networks?
>

[[rr]] A bipartite or two mode network is one with actors and receivers
which are logically different.. Multiple networks refers to more than one
relationship being included in the same model, with some effects defined
using both relationships. You can either fit an RSiena model with one
one-mode network or one bipartite network (this is the 'normal' case,
which can include behavioral variables and covariates) or include more
than one of the one-mode and bipartite networks, which we refer to as
'multiple networks'.

> With a 'nodeset', do you mean the type of networks (like for example
> adolescents and their friends (1) and romantic parners (2)) ?
>
 [[rr]] Actors and receivers are both examples of
what we call a node set in RSiena. The same node set may be a set of
actors for one network and receivers for another. They identify the nodes
among multiple objects in the model
> And why are you creating a network nodeset for the behavioral variable?
>
>
> and is the same or something different as a 'onemode' network?
>
[[rr]] If more than one set of nodes exists in your model I need to know
to which one the behavior variable refers. The nodeset is just a name for
a collection of nodes. Similarly for covariates I would need a node set
name to tell me which set of actors the covariates related to.

> and what do you mean with 'actors' and 'students' in your example? are
> they referring to the two types of networks (like adolescents' friends
> = actors and adolescent's romantic partners = students)?
>

[[rr]]Just names of nodes, with no deep meaning. Simply to identify them
as different. Logically I just need to know which nodes are which.

> And: don't you first have to create the nodesets, as in
>
> >
> > students <- sienaNodeSet(50, 'student')
> > Actors <- sienaNodeSet(50, 'Actors')
>
> and then use them in the bipartitie network:
>
> > mynet3 <- sienaNet(array(c(s501, s502, s503), dim=c(50, 50, 3)),
> > type='bipartite',  nodeSet=c('Actors', 'student'))
>
>
> Or is the logic different here?
>
[[rr]] Well, when you create a network object it just checks the number is
correct and stores the name(s) of the node sets. Not till sienaDataCreate
does it match them up and validate that they exist. So the order of
creation is not important.

Hope this helps,

Regards,

Ruth
 >
> Thanks for your time & thoughts!
>
> Maarten
>
> On Apr 23, 2010, at 14:38 , Ruth M. Ripley wrote:
>
> > Dear Maarten,
> >
> > Just add all the objects to one call to sienaDataCreate. The only
> > complication is if you have more than one set of node sets, when you
> > need
> > to specify any non-default one in the calls to sienaNet etc. And
> > create
> > them using sienaNodeSet eg
> >
> > for onemode network:
> >
> > mynet1 <- sienaNet(array(c(s501, s502, s503), dim=c(50, 50, 3)),
> > nodeSet='student')
> >
> > for behavior variable:
> >
> > mynet2 <- sienaNet(s50a, type='behavior', nodeSet='student')
> >
> > for bipartite network:
> >
> > mynet3 <- sienaNet(array(c(s501, s502, s503), dim=c(50, 50, 3)),
> > type='bipartite',  nodeSet=c('Actors', 'student'))
> > (this is a fake!)
> >
> > create the node sets: number of nodes and the name.
> >
> > students <- sienaNodeSet(50, 'student')
> > Actors <- sienaNodeSet(50, 'Actors')
> >
> > then create the data by just listing all the objects:
> >
> > mydata <- sienaDataCreate(mynet1, mynet2, mynet3,
> > nodeSets=list(students, Actors))
> >
> > You do not need to specify the node set name 'Actors' when creating an
> > object, other than for a bipartite network, but if you have any
> > other node
> > sets you do need to create the Actors node set and add it to the
> > nodeSets
> > list as here.
> >
> > Hope this helps, I obviously need to add some more examples to the
> > manual!
> >
> > Regards,
> >
> > Ruth
> >
> > On Wed, 21 Apr 2010, Maarten Van Zalk wrote:
> >
> >> Dear Ruth,
> >>
> >> I was wondering whether you could help me out with running multiple
> >> networks in RSiena. I understand how to make separate objects for
> >> each
> >> network, but how do I combine the objects for the analysis?
> >>
> >> I have data of around 850 adolescents and their friends inside school
> >> and friends outside school (2 networks). To start off, I was thinking
> >> about selecting just one class, and making a network of their friends
> >> within the same school and outside their school.
> >>
> >> Perhaps there is literature out there available on how to do this,
> >> but
> >> I have not been able to find it in the manual (only information on
> >> the
> >> formulae for the multiple network effects)...
> >>
> >> Thanks once again!
> >>
> >> Maarten
> >>
> >
>
>