***** To join INSNA, visit http://www.sfu.ca/~insna/ *****
Mark, it's true that the interactive bit with NetDraw that I reported is
different from yours in that it does not recalculate edge betweenness at
each step. Worth doing though.
On the other hand, recent* versions of Ucinet DO in fact do the
recalculation of edge betweenness after each edge deletion and produce
hierarchical partition tree. Note that I have only glanced at your article
(and that was a year ago), so it may still not be exactly what you've done.
I wrote the routine for my own use some time ago -- just dumped it into
ucinet now that there is a market for it.
Note also that my routine is not really suitable for large networks. Takes
over a minute for 200 node network of typical density. It calculates
edgebetweenness via canned routine, finds max value, deletes all edges with
that value, calculates components via canned routine, repeats. No shortcuts.
Presumably you've done some work on optimizing -- if you send me source code
*recent means the version I uploaded earlier today!
----- Original Message -----
From: "Mark Newman" <[log in to unmask]>
To: <[log in to unmask]>
Sent: Monday, March 31, 2003 11:48 AM
Subject: Re: Email networks and communities
> ***** To join INSNA, visit http://www.sfu.ca/~insna/ *****
> On Sun, 2003-03-30 at 18:48, Steve Borgatti wrote:
> > ***** To join INSNA, visit http://www.sfu.ca/~insna/ *****
> > BTW, this is kind of fun to play with interactively using NetDraw. In
> > Ucinet, run Edge Betweenness. Then in NetDraw, open the edgebetweenness
> > matrix as a network. Then in the Tie window, set the dichotomizing
> > to LT and the cutoff to the largest edge betweenness value. Then
> > press the "-" button to systematically lower the cutoff. As you do, you
> > components emerge.
> > steve.
> My experience with UCInet is not great, but I believe the calculation
> you describe is not the same as our algorithm, and our studies suggest
> that in most cases it will not give useful results. It appears to miss
> out the step in which the edge betweennesses are recalculated following
> the removal of each edge. This step is crucial to the generation of
> meaningful structure.
> As far as I am aware, neither UCInet nor any of the other
> general-purpose SNA programs can in their present form be used to
> implement our method (other than laboriously performing the
> recalculation step by hand). Of course, it would be possible to add
> code to UCInet to do so, and this might be sensible, seeing as the
> method has been used by quite a number of people since its first
> publication. I'm assuming each group implemented their own version - at
> least they didn't ask Michelle Girvan and me for our code - and this
> seems like an unnecessary duplication of effort...
> Best wishes,
> Mark Newman.
> 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.
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.