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

Figure 10

Scree plot, in logarithmic vertical scale, calculated from the correlation matrix, , for 3 network topologies: Erdös-Rényi (triangular orange markers), heterogeneous (circular blue markers), and Newman modular (squared green markers) networks, with parameters as in Figure 9. A scree plot shows the value of eigenvalues, , in descending order, as components (principal orthogonal directions) are gradually being included (up to the number of variables). is labelled in red, as a possible selection criterion for the number of components or factors. Other used criteria: much larger than the others and describing rather straight angles with successive values. 1-factor model and 4-factor model may be suitable for ER and Newman modular networks, respectively, whereas a more complex model is needed for heterogeneous network.