***** To join INSNA, visit http://www.insna.org *****
Hi Balazs -
I agree w. David that you need to start with a definition of cohesion
and then ask how that changes over time.
Doug White & I propose using node-connectivity as a measure of
structural cohesion. It has some nicely intuitive properties that
capture much of what theorists often mean by structural cohesion
(see: Moody, James and Douglas R. White. 2003.
Cohesion and Embeddedness." American Sociological Review
68:103-127). If you want to then treat "dynamic" as successive (or
overlapping) time slices of your network, then you can simply
calculate the cohesion for each time frame and map it's change in
levels over time.
However, a key feature of this notion of cohesion is that it focuses
on the node-independent path structure of the graph -- the multiple
ways that people can reach each other through the network without
that reach being uniquely controlled by a single (or small number of
) actor(s). An interesting aspect of a truly dynamic graph is that
the ordering of relations will shape flow through the
network. Effectively, all "flow" has to go downstream in time, with
local "reversals" possible when you have concurrent
relations. Since it is possible to identify these time-ordered
paths (Moody, James. 2002.
Importance of Relationship Timing for Diffusion: Indirect
Connectivity and STD Infection Risk" Social Forces 81:25-56), in
principle you could define a "common history of cohesion" through the
pattern of time-specific reachability (though I know of no
implementations of this, yet).
If you wanted to map the changing patterns of coalitions over time,
you might imagine treating "groups" as a class of nodes in a dynamic
two-mode network. At each time slice (whatever time is relevant
here), you assign nodes to "peer groups" (again based on your
context-relevant definition). You could then look at the one-mode
(presumably person) projection of this graph, which would effectively
show you the history of group co-membership over the life of your
network. The resulting structure of this graph might be very
informative: if you imagine people cycling though many groups then
the resulting projection will probably admit to few factions, if, on
the other hand, time-local groups are really embedded in a larger
constant coalition of units, this will come out in the
group-co-membership network as clear divisions.
And, of course, if you want to simply explore the dynamics of your
graph visually, you might want to check out SoNIA:
All the best,
At 02:33 AM 12/20/2005, you wrote:
>***** To join INSNA, visit http://www.insna.org *****
>I wonder if anyone had experience with identifying cohesive groups in
>dynamic network data.
>How would you identify a group as a common history of cohesion rather than a
>cohesive subset in a network snapshot?
>Any comments are appreciated,
>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.
Department of Sociology
Ohio State University
372 Bricker Hall
190 N. Oval Mall
Columbus, OH 43210
[log in to unmask]
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.