Weighted degree computation fails (all values 0) when a filter is applied #636

Closed
suthers opened this Issue Jun 2, 2012 · 7 comments

Projects

None yet

4 participants

@suthers
suthers commented Jun 2, 2012

If a filter is in effect when one computes "Avg. Weighted Degree" in the Statistics panel, all degrees are set to 0.0 for entities that pass the filter. (Previously computed weighted degrees for entities not passing the filter correctly are unaffected.)

Thus I cannot, for example, compute the weighted degree of a node within its modularity class. I am forced to use the global weights, computed without any filters in effect.

@suthers
suthers commented Jun 2, 2012

BTW this is a serious bug for us as we are doing a data analysis this week for a paper to be written next week and due June 15th and the weighted degrees represent level of activity of the actors being considered in the analysis. Due to this bug we cannot say how much an actor is active in each candidate community offered by the modularity partitioning, and will have to report this as a limitation of Gephi in the paper. A fast fix would be greatly appreciated!!!

@suthers
suthers commented Jun 2, 2012

Sorry for not noting this in the original post but further info:

  • unweighted degrees are correctly computed when filters are applied,
  • I verified the bug with weighted degree using a different partition filter (it is not just modularity classes) and using a degree range filter (it is not just partitions).
    Thanks -- Dan
@eduramiba
Member

Hi,
As a workaround while this is fixed, I think you could export your filtered graph to a new workspace and run stastistics there.

@suthers
suthers commented Jun 3, 2012

eduramiba, thanks for the suggestion. The problem is that I am working with 234 modularity classes and need the weighted degree computation for each of them. So I would have to export 234 graphs and reload each as a project. Not sure I can do this by the paper deadline of June 15th, let alone write the paper, if I take that approach!

@sheymann
Member

You may export the Node Table and Edge Table to compute the statistics in R. This is what I would do.

@mbastian mbastian was assigned Jul 7, 2012
@mbastian mbastian added a commit to mbastian/gephi that referenced this issue Jul 7, 2012
@mbastian mbastian Fix issue #636 80fb770
@mbastian
Member
mbastian commented Jul 7, 2012

Problem fixed. Thanks for the report hopefully you found a way for your paper. You may want to write a plugin to compute the avg degrees for all 234 partitions automatically.

@mbastian mbastian closed this Jul 7, 2012
@suthers
suthers commented Jul 12, 2012

Thanks Mathieu. We have another plugin project underway, to implement
the Ahn et al. link community algorithm (overlapping node
communities). The Blondel et al. algorithm in Gephi is good but
nonoverlapping. The popular Palla et al. clique percolation for
overlapping communities does not work on bipartite graphs!

--On July 7, 2012 4:45:28 PM -0700 Mathieu Bastian
reply@reply.github.com
wrote:

Problem fixed. Thanks for the report hopefully you found a way for
your paper. You may want to write a plugin to compute the avg
degrees for all 234 partitions automatically.


Reply to this email directly or view it on GitHub:
#636 (comment)


Dan Suthers
http://lilt.ics.hawaii.edu/ http://www2.hawaii.edu/~suthers/
Email can take a week or more to read: for urgent matters please
call 1-808-956-3890 (office), text 1-808-741-5686 (mobile), or
drop by my office in POST 309.

@mbastian mbastian added Fix Released and removed Fix Committed labels Nov 21, 2015
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment