This is a brief summary on comparison of brain connectomes at multiple scales, as reviewed in [Meskaldji, NeuroImage, 2013]. We focus here on brain connectomes, but the concepts can also apply to other complex networks.
A connectome is a network model of the brain architecture. Given the enormous number of anatomical connections, connectomes easely become very complex, usually containing up to thousands of nodes and connections. When comparing connectomes between populations, appropriate statistical procedures must be taken. This is specially the case when differences at the level of subnetworks or single connections are inspected, as the consciderable number of tests that are performed requires the application of correction methods in order to control the rate of false discoveries.
The analysis of connectomes can be done at three different scales : global, subnetwork and local