The “Raw Matrix (link wise)” panel offers to perform network based statistics
(association matrix based statistics) by applying either correlational and/or group
comparison analyses (please refer to the “Raw Matrix –Correlation”
and “Raw Matrix – Group” section below). The addition “link wise” means that the respective analysis is
performed on each link between any two nodes in the n x n matrix. Thus, this feature is
inherently a mass-univariate approach that examines the significance of each link with
respect to the applied analysis. However, to deal with the problem of alpha inflation
(i.e., multiple comparison problem), GraphVar makes use of so called Graph-Components
(i.e., subnetworks in which all pairs of nodes are connected by significant links) that are identified
by the BCT “get_components” function. The measure of interest with regard to Graph-Components
is their size (this is similar to a contiguous cluster of voxels in fMRI).
To evaluate, whether a Graph-Component with a certain size as a result of the respective
analysis is non-random, it is compared against the amount and size of Graph-Components in
“random” data (please refer to the respective analysis type described below). Based on this,
a p-value for each non-random Graph-Component is computed similar as in AFNIs AlphaSim
(for how to view the resulting Graph-Components and a more detailed description of the p-value
computation for Graph-Components please refer to the “Network Inspector” section).


TIP: with the raw matrix function, it is also possible to examine correlations/group
differences between a set of a-priori specified structures (e.g. correlation of age on
insula-amygdala connectivity). For this function you may select the specific structures
in the “Network nodes/Brain areas” selection window. When working
with the entire matrix, you can also select specific regions using the “Results viewer”
in the “The results selection box”.