Fig. 1: Sketch: for each population matrix, we estimated for each element (here exemplified by element k3,2) an integrated sensitivity with respect to entropy (∆Hk3,2) and with respect to fitness (∆ λ k3,2) by increasing (perturbing) element k3,2 by amount b and simultaneously reducing elements k1,2; k2,2; k4,2 by b/n with n=1-(Number of non-zero column elements). Such integrated sensitivities were then computed for each matrix element ki,j and for both types of sensitivities. For each population matrix we fitted a linear model through data points based on these two types of sensitivities from each of the ki,j elements. Each line in Fig. 2 corresponds to such a correlation model.