Fig. 3
From: Path-level interpretation of Gaussian graphical models using the pair-path subscore
A schematic showing how different paths in a GGM can contribute to the correlation between two nodes a and b. Panel a shows the paths connecting a and b, while panel b shows the same paths with possible PPS values (note that these sum to 1). From PPS, we learn that the edge between a and b, and the path through node d, are the largest contributors to the marginal correlation between nodes a and b