A new version of SocNetV has just been released! Version 1.5 brings standardization, many bugfixes and nice new features (see below). Packages for Windows, Mac OS X and Linux are available from the Downloads page.
Notable among the new features are:
- Prominence Scores on Valued Networks:
A new SSSP-solver algorithm has been implemented (Dijkstra) which allows SocNetV to compute prominence indices on weighted networks. When the graph edges are weighted, the application asks the user if it should consider weights in computations. Also, it asks the user if weights should be inverted or not. This question is crucial since edge weights can have different meanings. For instance they can denote cost or votes. If they denote cost, then the geodesics should be those paths with minimum value. But, if the weights denote votes, then the geodesics could be those path with maximum value. In the latter case, the user should choose to invert weights so that Dijkstra compute the desired paths and distances.
- Standardized Centrality and Prestige scores:
From this version all actor prominence indices report standardized scores (from 0.0 to 1.0) where applicable. If there is no known formula to compute such a standardized score for actors, then SocNetV computes a std score by dividing the original index score by the sum of index scores of all actors. For instance, this happens on PRP. Warning: If the original prominence index has range from 0 to 1 (i.e. EC and IRCC), SocNetV considers that as std and does not compute anything else.
- Prominence layouts are relative to highest score
From this version, all (circular, level and nodal size) visualization layouts based on prominence scores are graphed relative to the highest score in the network, instead to the theoretical max (1.0). For instance in a circular layout, say the actor with the highest CC has score 0.8. That node will appear to the center of the circular layout. All other actors will appear on circles of radius relative to that highest score. I.e. an actor with score 0.4 will appear on a circle of radius 50% further from the screen center, while another actor with CC score 0.2 will appear on a circle of radius 75% further from the actor with the highest score.
- Skip isolates, if you like
The user can omit isolates and compute prominence indices for the resulting graph.
- Graph connectedness
SocNetV can report the network connectedness (whether it is a connected graph or digraph, unilateral etc). Also it can check whether isolates exist that can be removed so that the graph can become connected
- CC drops isolates by default
Up to v1.4, SocNetV did not compute CC scores if the network had isolate nodes, instead it urged the user to use IRCC. From this version, SocNetV checks if isolates exists and automatically drops them in order to compute CC scores.
- New datasets
- Stephenson and Zelen (1989): Network of 40 AIDS patients
- Stephenson and Zelen (1989): IC test dataset, 5 actors
- Wasserman and Faust: star, circle and line graphs of 7 actors
Also, the following bugs have been fixed:
- #1358678 Fix GDC calculation in weighted networks
- #379558 Force-Directed algorithms produce poor layouts
- #1365028 Methods isOutLinked & setOutLinked do not consider relations
- #1365504 centralityInformation() should symmetrize adjacency matrix
- #1366625 click on an edge does not select the right edge
- #1369171 Group IC calculation yields incorrect results
- #1371208 wrong power centrality scores
- #1364955 vertices() should report only enabled vertices
- #1369336 pagerank prestige reports wrong scores
- #1370528 socnetv cannot build on non-x86-based architectures
- #1364320 remove SRS pdf and fix spelling errors in Code
- #1364361 SocNetV does not remember last directory used by user
- #1378346 Cannot change size and value of a node