(A) Firstly, the full dataset is decomposed into a number of components/modes (the generic outer product model shown here is relevant to both ICA and PROFUMO). (B) Full Pearson’s correlation matrices are estimated from the elements shown in A (the subscript s refers to matrices that have been standardised over time for Y and T, and standardised over space for S; n=#TRs for the dense connectome and the temporal network matrix, and n=#grayordinates for the spatial overlap matrix). The correlation matrix of the full dataset (Y) results in a voxel-to-voxel matrix referred to as the ‘dense connectome’. The correlation matrices of the component maps or timeseries result in component-to-component matrices. Here, each element in the temporal network matrix describes the correlation between two timecourses (i.e., a temporal edge), whereas each element in the spatial overlap matrix describes the correlation between two spatial maps (i.e., a spatial edge).