Identified regions cluster into two functional networks, which show decreased across-network correlation in schizophrenia. Functional interrelationships among the detected regions, as represented in the region-by-region correlation matrices pooled across groups, were analysed by ( A ) principal component analysis (PCA), with a scatterplot constructed for the first two principal components for each region’s pattern of correlation. Analogous PCA results calculated on the corresponding group t -test matrix are shown in B . The results of K-means cluster analysis on the average correlation matrix (pooling groups) are shown using colour (magenta and green, matching Clusters 1 and 2 in Fig. 1 , respectively) in A and B . An ‘elbow’ plot constructed for K-Means cluster analysis ( C ) shows that the trade-off of variance explained versus the complexity of the cluster model (choice of K) is optimized at K = 2 clusters. Region-by-region correlation matrices sorted by cluster membership are then shown in D for controls, E for patients with schizophrenia, with the corresponding t -tests (Control-Schizophrenia) shown in F , having removed nuisance covariates of Age and Motion. The row/column order of regions of interest in the matrices is identical to those provided in Table 2 (regions 1–26 from bottom to top row and left to right column).