I have few questions that I hope you can help me with:
1. Should there be put a threshold on weighted graphs? Some studies have done that while other use the weights (connectivity values) directly.
2. When calculating path length and clustering coefficient in undirected weighted networks there seems to be a strong inverse correlation between the two if no threshold has been applied. Is this true?
2. I haven’t been able to find this out: Is there any reasons for using Dijkstra algorithm or Floyd–Warshall algorithm when calculating average path length in undirected weighted networks?
1. It’s not necessary to threshold weighted graphs, but it can be done of course.
2. Yes, there is a correlation. In general, there is a correlation between all measures and the average weight of the graph. Avoiding this issue is one of the man arguments in favour of using binarized matrices instead of the weighted ones.
3. There is no specific reason. Any (correct) algorithm should give you the same result.