Figure 2:

Example causal diagram for (a) GBM survival prediction and (b) the same causal diagram of GBM with links and nodes representing expected confounding information and population differences for variables. In the diagram, solid circular nodes represent observed variables; while square nodes indicate selection nodes controlling for population differences. Causal links are represented with solid lines with directional arrows. Bi-directional dashed lines indicate a variables linked by confounders. The selected observational variables are Patient Age (Age); Karnofsky Performance Score (KPS); 9-gene metagene score (Metagene); and Patient Survival at Population Median (Survival). Unique selection nodes for Age and KPS are shown as S_A and S_K.