A Complex Network Framework to Model Cognition: Unveiling Correlation Structures from Connectivity

Figure 12

Percentage of the total explained variance of correlation matrix , (29), for each of the first five components as a function of the edge probability, , for an Erdös-Rényi of size (a), and intracluster degree, , for a Newman modular network of size and (b). Both networks enable a factor model as a good descriptor of the outcome. Selected significant number of components are highlighted in orange and green colors, for ER and Newman modular networks, respectively. As nodes become more connected and communities more delimited, factor model moves from mirroring a noisy identity matrix to be a clear indicator of the correlation structure. First component (circular orange marker) increases as does so, strengthening the validity of a 1-factor model (a), while second to fourth components (circular, starry, and triangular green markers) increase as does so. Hence, an increase in modularity reinforces the validity of a 4-factor model (b). Parameters are set as in Figure 6, constraining the values to lay within the metric stable state regime.

(a)

(b)