GBC different in Pajek cp. to SocNetV

[Dimitris Kalamaras]

Thomas, a student from Germany reports:

"[...]But when it comes to calculating the Group Betweenness Centralisation, I am not getting the same result using the same input-file with Pajek and SocNetV.
When calculating the GBC, you mention the reference of Wasserman & Faust, but I am getting a wrong result.

Example:
--------------------------------------------
*Vertices 5
     1 "V1" 0.5 0.8 0.5000
     2 "V2" 0.8 0.4 0.5000
     3 "V3" 0.6 0.1 0.5000
     4 "V4" 0.25 0.6 0.5000
     5 "V5" 0.3 0.2 0.5000
*Arcs
 1 2 1
 2 3 1
 2 5 1
 3 5 1
 3 1 1
 4 2 1
 5 4 1
*Edges
---------------------------------------------
Using Pajek and calculating the GBC on my own, GBC=0,45833.
Using SocNetV: GBC=0,3125.

As far as I know are both programs using Wasserman & Faust, p.192, 5.14 for calculating the GBC.

I haven't found a mistake in my calculation, so can you have a look on my example and give me notice of what you found out."

Report also here:

http://sourceforge.net/forum/message.php?msg_id=7482599

[Dimitris Kalamaras (dimitris-kalamaras)]

There was a bug in the last section (accumulation) of Graph::createDistanceMatrix() method: SocNetV was using a wrong maximum BC value. Thanks to Thomas' report, it has been fixed:
http://socnetv.svn.sourceforge.net/viewvc/socnetv?view=rev&revision=367