Is there any statistic that can tell how "hubby" is a graph?
By this I mean a number that is small when a graph has no hubs, that is, when all nodes are more or less equal degree-wise, and big when there are hubs, nodes that concentrate most of the connectivity.
I expect it to be zero (or minimal) for a fully connected graph and 1 (or maximal) for a "star shaped" graph with one node with degree N-1 and N-1 nodes with degree 1.
I was thinking to use some measure of the dispersion of the degree distribution (degree entropy, degree variance, ...). Or maybe to define a hub as a node that, when cut out, turns a connected graph in a non-connected graph and define its "hubbiness" as the number of connected components after I cut it from the graph (and define the hubbiness of the graph as the average hubbiness of the nodes).
But I'm not a specialist in this field, so I don't know if there is an already tried and established measure in the specific community, easy to estimate.
my graphs are usually small (in the worst case 100 graphs with 100 nodes each, but typically 20 graphs with 20 nodes each), so I'm slightly (not absolutely) more interested in theoretical soundness than efficiency.