***** To join INSNA, visit http://www.sfu.ca/~insna/ *****
One way to look at clustering methods is as models of group structure.
Classifying actors into mutually exclusive groups (i.e., imposing a
partition) can yield a pretty distorted view of the network -- that is,
be a bad model. Which is why one wants a measure of fit. Even if the fit
is not perfect has the advantage of simplicity. It is, in the words of
Levi-Strauss, "good to think with". As long as we remember that it is a
model, not reality.
I think clustering methods (whether they produce partitions or
overlapping groups) are easily justified as ways to produce simplified
models of the structure (i.e. data reduction), but it requires something
of a leap of a faith to see them as revealing emically defined groups,
by which I mean groups that the actors themselves believe to exist and
identify themselves as members of. [I apologize in advance for that last
> -----Original Message-----
> From: Social Networks Discussion Forum [mailto:[log in to unmask]]
> On Behalf Of Valdis Krebs
> Sent: Monday, April 05, 2004 2:33 PM
> To: [log in to unmask]
> Subject: Re: [UCINET] friendship groups
> ***** To join INSNA, visit http://www.sfu.ca/~insna/ *****
> Of all of the social networks I have looked at, I can not remember a
> single one where everyone fell into one and only one group/cluster.
> You will find 'clean components' if you are looking at a *prescribed*
> network such as a department structure in an organization or a class
> structure in a school but NOT with most *emergent* networks.
> Are you interested in how the network influences individual behavior?
> If so, I would look at the network neighborhood each individual is
> embedded in.
> If you are just looking to discover emergent friendship clusters in
> the overall network, I would apply Mark Newman's recent edge
> betweenness algorithm
> SOCNET is a service of INSNA, the professional association for social
> network researchers (http://www.sfu.ca/~insna/). To unsubscribe, send
> an email message to [log in to unmask] containing the line
> UNSUBSCRIBE SOCNET in the body of the message.
Organization Studies Dept
Fax: 978 456 7373
Tel: 617 552 0450
E-mail: [log in to unmask]
SOCNET is a service of INSNA, the professional association for social
network researchers (http://www.sfu.ca/~insna/). To unsubscribe, send
an email message to [log in to unmask] containing the line
UNSUBSCRIBE SOCNET in the body of the message.