Figure 3
From: Network evaluation from the consistency of the graph structure with the measured data
Robustness regarding graph structure variation. The calculation is composed of five steps (see details in the text). Three networks with typical structures in biology are examined in (A), (B), and (C). To generate the simulation data by structural equations, we set the standard deviation to 0.1 in all three graphs, and the path coefficients between the variables are as follows: (A) α1,2 = 0.6, α2,3 = 0.3, α3,4 = 0.1, α4,5 = 0.7, α5,6 = 0.8, α6,7 = 0.9, α7,8 = 0.2, α8,9 = 0.5, and α9,10 = 0.4; (B) α1,2 = 0.1, α1,3 = 0.2, α1,4 = 0.3, α1,5 = 0.4, α1,6 = 0.5, α1,7 = 0.6, α1,8 = 0.7, α1,9 = 0.8, and α1,10 = 0.9; and (C) α1,2 = 0.5, α1,3 = 0.7, α2,4 = 0.4, α2,5 = 0.8, α3,6 = 0.6, α3,7 = 0.3, α4,8 = 0.2, α5,9 = 0.1, α6,9 = 1.0, and α7,10 = 0.9. The value of log-likelihood and the parameters of GEV distribution in the respective networks are as follows: (A) l(G0) = 163.4805, μ = 89.8375, σ = 12.9694, and ξ = -0.1743; (B) l(G0) = 61.6096, μ = 3.0217, σ = 12.5220, and ξ = -0.1314; and (C) l(G0) = 124.8894, μ = 46.9002, σ = 12.1395, and ξ = -0.1406. See also the corresponding GEV plots at additional file 5: Robustness regarding the network structure variation.