Hello socnetters,

I have been thinking a lot about spring-embedders as a way not only to
visualize structural data but also as a way of actually fetching information
on centrality and subgroups/cliques. After experimenting a bit with
spring-embedders in open-space...
(2-dimensional, open-space)
(3-dimensional, open-space)

...I intuitively feel that these are inadequate to identify subgroupings
between actors which are relative peripheral/non-central. These
spring-embedders all have a natural central point and the most dominant
actors and groups tend to occupy this center and disturb the placements of
more peripheral, weaker actors and their ties to eachother.

By using a closed-space setup, I do believe that this 'gravity problem' can
somewhat be avoided. I have developed such a closed-space spring-embedder to
test this hypothesis - this program (run as a java applet directly through a
web browser) and instructions on how to experiment it is to be found here:

The comparative analysis I have done so far is by using the hierarchical
clustering algorithm which is to be found in Ucinet. The spherical,
closed-space scenario seems to be better at visualizing non-central
groupings than a flat, open-space scenario does. Three dendrograms - one for
the original data set and two containing geometric distances for the open-
and closed-space spring-embedders - are available at the above URL.

I would greatly appreciate any feedback on this as I am not sure how to
interpret the result. A spring-embedder can probably never be perfect at
identifying subgroups but, intuitively, they should be quite good at
visualizing/finding a combination of subgroupings and centrality/prestige.
But how should I continue analysing the geometric datasets obtained from the

Yours very sincerely,
Carl Nordlund
Carl Nordlund, BA, PhD student
Human Ecology Division