***** To join INSNA, visit http://www.insna.org *****
this problem is very interesting. We are working on similar topic and
so far we have experimented different approaches because we are still
not sure about the best way to proceed, instead of comparing directly
the centrality values (that may suffer from distribution problems and
also normalization and network-size problems when we compare different
networks) we have got interesting results computing the rankings and
comparing them (e.g., x axis: rank in betweenness centrality in FF
net, y axis: rank in betweenness centrality in RT network).
WRT to removing nodes that did not interact, it probably depends on
what you want to mine from the data, but we tried to consider them in
our analyses whenever possible - intuitively, not considering nodes in
one network that are not present in the other(s) may reduce the
differences found by the mining methods. If one node is the most
central in the FF network and did not interact at all with the others,
this is very relevant information.
You might find something interesting in our recent work (some
references of recently published stuff on it may be ICWSM 2012 paper
by Rossi and Magnani on influence between FF and interaction networks
in Twitter and, more generally ASONAM 2011 paper by Magnani and Rossi
on multi-layer network models - we will also soon put something
related on Arxiv).
Most of it should be linked in our website: http://larica.uniurb.it/sigsna
Hope this can be useful.
On Fri, Jul 27, 2012 at 11:52 AM, Thomas Plotkowiak <[log in to unmask]> wrote:
> ***** To join INSNA, visit http://www.insna.org *****
> I have the following problem:
> 100 people on Twitter. And 3 Networks between them: Friend & Follower
> (FF) connections, AT-Interactions (edge A-->B ~ A writes @B ),
> RT-Interactions (edge A-->B A retweets one of B's tweets).
> I want to find out how much the centrality of a person in the FF and
> AT network influences the centrality in the RT network (aka. people
> with high centrality in the RT network are people that have been
> retweeted often)
> Now I have 2 problems:
> Problem 1:
> It turns out that there are people that have FF connections but don't
> get mentioned at all and don't get retweeted, therefore their
> in-degrees and centrality metrics in the AT and RT networks are 0.
> Should I exclude these people from the analysis?
> The argument for doing this is similar to the assumptions that have to
> hold for e.g. SIENA when modeling network evolution. Here actors also
> have to be existing in all panels.
> Problem 2:
> Is it feasible to to perform such a regression given the nature of the
> log-normal or powerlaw distribution of in-degrees or centrality
> Bonus question:
> If I were to model these correlation of these networks in pNet and try
> to capture some sort of centrality effect, is that possible? Would it
> simply be incoming-network ties effect or someting like that?
> Best Regards
> Thomas Plotkowiak
> Research Assistant MCM Institute
> 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.
Un contributo importante al finanziamento delle attività di ricerca.
5x1000 all’Università di Urbino. C.F.: 8 2 0 0 2 8 5 0 4 1 8
LaRiCA - Advanced Communication Laboratory
Faculty of Sociology - "Carlo Bo" University, Urbino
[log in to unmask]
T. +39 0722 305726 F. +39 0722 305727
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.