***** To join INSNA, visit http://www.insna.org *****
<<<-------- Carl Nordlund-------->>>
> ***** To join INSNA, visit http://www.insna.org *****
>
> With all due respect, and very grateful for your help, I do still think
> that there is some problems with the indices in the formula you supplied
> in your email.
The main problem in the construction of (dis)similarity measures
using a matrix of some relation between vertices A[u,v] is that
computing the dissimilarity d(u,v) we can't simply compare the
row A[u,.] with the row A[v,.] (and/or the column A[.,u] with the
column A[.,v]) - we can do this for vertices different from u and v;
but we have to compare A[u,u] to A[v,v] and A[u,v] to A[v,u].
If we are using in our measure both rows and columns then we are
making this comparison twice - in some measures there is a special
parameter specifying how many times we consider these comparisons
(zero, once or twice).
See Batagelj V., Ferligoj A., Doreian P.: Direct and Indirect
Methods for Structural Equivalence. Social Networks 14(1992),
63-90 or Patrick Doreian, Vladimir Batagelj, Anuška Ferligoj:
Generalized blockmodeling, CUP 2005, page 178-184.
Vlado
--
Vladimir Batagelj, University of Ljubljana, FMF, Department of Mathematics
Jadranska 19, 1000 Ljubljana, Slovenia
http://vlado.fmf.uni-lj.si
_____________________________________________________________________
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.
|