This section really needs a major rewrite, and I'm not sure that leaving it in as presently constituted would be a good thing. The scores returned by the two libraries for the imf2014 dataset are wildly different, and do not resemble each other at all, and the exercise to "find the top 5 eigenvector scores" is pretty meaningless as a result.
We either need to recommend not looking at eigenvector scores for directed or weighted networks (and the negative weights in the imf2014 dataset might be affecting things as well), or really get to the bottom of why the scores are so completely different from each other, and offer some guidance on which library to use if differences cannot be resolved.
This section really needs a major rewrite, and I'm not sure that leaving it in as presently constituted would be a good thing. The scores returned by the two libraries for the
imf2014
dataset are wildly different, and do not resemble each other at all, and the exercise to "find the top 5 eigenvector scores" is pretty meaningless as a result.We either need to recommend not looking at eigenvector scores for directed or weighted networks (and the negative weights in the
imf2014
dataset might be affecting things as well), or really get to the bottom of why the scores are so completely different from each other, and offer some guidance on which library to use if differences cannot be resolved.