Doug, you might look at my book, Social Semigroups, to see how I handled
the positive and negative relationships in the Sampson data. E.g., if you
look at the correlations between compositions of relations (using real,
not boolean, matrix multiplication) you can get a small semigroup. With a
cutoff of .8 for compositions up to length 4 I got an eight element
semigroup with such multiplications such as interesting equations as
aaa = aa,
bbb = ab,
bab = bb,
etc, where a is the "friendly" relation and b is the "unfriendly".
John P. Boyd
Department of Anthropology
University of California at
Irvine, CA 92697-5100
On Tue, 23 Apr 2002, Doug Bryan wrote:
> I have egocentric network data with ties that are labeled as "friend" or
> "foe". There are a number of kinds of analysis I can do on the subnet
> (subset of the data) containing only friend ties, and similarly on the
> foe-only subnet. But are there methods that work on the network that
> contains both kinds of ties?
> References and pointers would be greatly appreciated.
> Doug Bryan
> [log in to unmask]