Revised version of a presentation at the 3rd European Conference for
Network Analysis, Muenchen, 1993.
Thanks to Cynthia Lehmann for revising my english manuscript. More information about network visualization is available via Mosaic
under following URL: http://www.mpi-fg-koeln.mgp.de/lk/netvis.html
- The size of the nodes in this and all following figures
is proportional to the sum of both the in- and outdegree of a node.
This characterizes a node by all direct activities it is engaged in
and reflects the
direct component of centrality.
- The two different
rank-orders of centrality are listed in Table : for
direct links the algorithm proceeds according to
the rank-order given in the column
'Degree of Centrality', while for geodesics the rank-order given under 'Closeness' is used.
As the reader can verify, the 'boundary spanners' have higher ranks (are more central)
when geodesics are used.
prevent unwanted weights from entering into the optimization procedure,
the algorithm locates
each of the organizational shapes on a second circle,
choosing the diameters of the larger circle to correspond to the
diameters of the shapes representing the single systems.
While Figure was one of the first results we achieved
with our algorithm, there is also some further criticism because circles
C and D seem to be interchanged. This results from using a square
arrangement for the four circles. Because the diagonal whithin a square
is longer than one of its sides,
the overall distance for the total system
could have been reduced by interchanging C and D. For an
automatic solution, we would have to use a two-level strategy:
first to assign each of the four a priori blocks to their best
fitting shape, and then to rearrange the nodes in each of the
subsystems in a second step.
Fri Mar 31 13:14:02 MET DST 1995