***** To join INSNA, visit http://www.insna.org *****
You've got two problems, two deep problems:
First is cognitive style. It should not affect what we produce as theory, but
it does. One set of people is at ease with discete objects: with sets, with
relations described by a table, with abstract algebras performed on those tables
(interpreted as matrices). White was demonstrating that mode with CONCOR. Another
set of people is at ease withcontinua and variable, with analysis and conventional
linear algebras. MDS comes out of this 'style'.
De facto you will often find that the first CONCOR split will correspond to a
line drawn across an MDS solution. De facto you will often find that the second
CONCOR split will correspond to another line (or to two not quite connected lines)
that divide the two regions established by the first line.
Second problem, there is no such thing as THE solution by MDS. If I take a picture
of my house, we can pretty well judge whether it is a good picture. If I compute
an image of a network -- who knows? That's quite another matter (no reference
point, no eyeball to look at the object directly and then compare it to the MDS
image.
You will find that different scaling procedures produce different maps -- a reality
we normally try to make go away by using only one set of software. That does
not solve the problem.
Joel Levine
_____________________________________________________________________
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.
|