Fig. 8.

Density distribution of Fleiss Kappa statistics across 50 pathways for the non-null models when there’s no noise in the simulation. Fifty pathways from hallmark collection have been separately used in the non-null model workflows for total 100 runs. Fleiss Kappa was calculated to measure the agreement among these 100 runs. a Distribution of Fleiss Kappa in the first simulation study. Pathway-level model has better model stability than gene-level model. b Distribution of Fleiss Kappa in the second simulation which broke the inter-gene correlation. Without inter-gene correlation, pathway-level model became more advantageous in model stability compared with gene-level model