***** To join INSNA, visit http://www.insna.org *****
David gave an insightful reply mine may be redundant -
In my studies of classroom communication (which is more disorganized than
business meetings), there tend to be simultaneous streams of conversation
and the sequences intermingle. There are also lots of false starts, etc.
This can occur in group meetings or even on ebay chat groups (just follow
the strings there). In general, however, focused encounters (or controlled
ones) have one dominant strand with some byplay so I think David's approach
works really well. Moreover, it's more in line with conversation analysis'
Probably because of the chaos present in many high school classrooms, I've
taken a slightly different approach to studying participation frameworks in
the Goffman sense and relating them to networks. I've coded communication
with regard to their intended arenas: some talk is private (so dyadic
chains), other talk is localized and public (so bystanders are drawn in -
use seating chart to get at that), and then other talk is in the form of a
global broadcast statement (think gas to all participants). By rendering
these arenas of talk into network form (you can even weight co-authors vs
bystanders differently), you quickly identify different coordination games
or tasks and how people switch across / break them.
As for weights - David is again right that the coding influences various
measures. I've tried it weighted and unweighted (1 to all vs .1 to each of
10), and have tried omitting group directed remarks. In the end, I think
I've settled on the unweighted inclusion of everything because there are big
assumptions being made as to how information is received that I can't seem
to adequately justify to reviewers.
Couple ideas to play with - get duration information if you can, and also
consider that actor outdegree / energy may be captured by weighted data
while network paths / patterns/positions may be better captured by
Last - if you break aggregations down far enough, you end up with sequences.
On 5/24/05 10:01 AM, "David Gibson" <[log in to unmask]> wrote:
> ***** To join INSNA, visit http://www.insna.org *****
> If you're aiming to do a conventional analysis of the resulting network, you
> have to decide whether to treat a group-directed remark as entailing
> simultaneous communications from ego to each alter, or one communication from
> ego to a fictitious group actor (another column in the matrix -- though
> not one from which communications can issue), or a communication to no one at
> all. My preference has been to treat group-directed remarks and
> ambiguously-directed remarks identically, mainly because they can be difficult
> to distinguish in practice. Daniel McFarland, in contrast, prefers to think of
> a group-directed remark in terms of simultaneously activated dyadic channels.
> His solution is more compatible with conventional network measures and
> If you're just aggregating through time, you might divide by the number of
> people to whom an utterance was simultaneously directed, so that a direct
> remark from me to you would add 1 to the i,j cell, while a remark from me to a
> group of ten people (excluding myself) would add .1 to each i,j cell. Of
> course, interpretation will be tricky, for instance of centrality if you don't
> buy the implicit phenomenology behind that weighting. Another option would be
> to analyze the network of directed remarks and tendencies toward
> remarks separately.
> Whichever approach you take, be warned that if you're aggregating through time
> in order to generate one summary matrix of who spoke to whom, you're at risk
> eliding sequential dependencies, such as the fact that it's hard to address
> someone who didn't recently speak. In my view, time-collapsed who-to-whom
> matrices are partially artifacts of those kinds of sequential effects.
> Some references:
> Gibson, David R. 2005. "Taking Turns and Talking Ties: Network Structure and
> Conversational Sequences." American Journal of Sociology (forthcoming May
> issue, I hope).
> Moody, James, Daniel McFarland, and Skye Bender-deMoll. 2005. "Dynamic Network
> Visualization." American Journal of Sociology 110:1206-41.
> David Gibson
> David Gibson
> Assistant Professor
> Department of Sociology
> Harvard University
> 564 William James Hall
> 33 Kirkland Street
> Cambridge, MA 02138
> Voice: (617) 495-3825
> Fax: (617) 496-5794
> Quoting "Van Patten, Isaac T" <[log in to unmask]>:
>> ***** To join INSNA, visit http://www.insna.org *****
>> In analyzing the communications matrix of a group meeting (who addresses
>> whom) is it legitimate to include "Group" as a recipient of
>> communications addressed to the group as whole in a directional network?
> 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.
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.