Thanks to Kit Baum, a new program is up on SSC to compute Network's Weighted Clustering Coefficients.
to install.
Nw_wcc returns the clustering coefficient of the network's nodes (besides overall and average); following the indexes developed by Fagiolo, Phys. Rev. E (2007).
As in Fagiolo's article, five different "patterns" of clustering coefficients could be computed, and the desired pattern should be indicated in option. By default, the All (D) pattern is computed.
By default the weights are non normalized (as in Fagiolo, Phys. Rev. E (2007)), but a normalized option is added for weighted networks whose weights outpace unity.
Two normalization procedures could be indicated (max or sum normalization).
As discussed recently (here), without further indication, the normalization option would act like normalization(max).
A last option (binary) allows to compute the clustering coefficients ignoring the edge's weights (constructing binary directed networks).
Please note that nw_wcc requires T.Grund nwcommands package to work, available on github (also see further details in nw_wcc help file).
Also, this clustering coefficient module does not conflict with the previous nwcluster I've developed, because they simply don't compute the same clustering coefficients indexes:
nwcluster is based on Onnela et al, Phys. Rev. E (2007) weighted generalization of clustering coefficients.
nw_wcc follows Fagiolo, Phys. Rev. E (2007). and details several clustering patterns in directed networks.
Such that both versions are still available on SSC, the helpfile of each one pointing toward the other one and explicating the differences.
I hope this would be useful for all networks analysts in Stata.
Best,
Charlie
Code:
ssc install nw_wcc
Nw_wcc returns the clustering coefficient of the network's nodes (besides overall and average); following the indexes developed by Fagiolo, Phys. Rev. E (2007).
As in Fagiolo's article, five different "patterns" of clustering coefficients could be computed, and the desired pattern should be indicated in option. By default, the All (D) pattern is computed.
By default the weights are non normalized (as in Fagiolo, Phys. Rev. E (2007)), but a normalized option is added for weighted networks whose weights outpace unity.
Two normalization procedures could be indicated (max or sum normalization).
As discussed recently (here), without further indication, the normalization option would act like normalization(max).
A last option (binary) allows to compute the clustering coefficients ignoring the edge's weights (constructing binary directed networks).
Please note that nw_wcc requires T.Grund nwcommands package to work, available on github (also see further details in nw_wcc help file).
Also, this clustering coefficient module does not conflict with the previous nwcluster I've developed, because they simply don't compute the same clustering coefficients indexes:
nwcluster is based on Onnela et al, Phys. Rev. E (2007) weighted generalization of clustering coefficients.
nw_wcc follows Fagiolo, Phys. Rev. E (2007). and details several clustering patterns in directed networks.
Such that both versions are still available on SSC, the helpfile of each one pointing toward the other one and explicating the differences.
I hope this would be useful for all networks analysts in Stata.
Best,
Charlie
