Fig 1.
Modules with sale-free structures used in Simulation I.
The left panel is Module 1 and the right panel is Module 2. These modules were simulated based on Barabási–Albert model using package igraph in R.
Fig 2.
Modules with non-scale-free structures used in Simulation II.
The left panel is Module 1 and the right panel is Module 2.
Table 1.
Type I error rates averaged over 1000 replicate data sets.
Fig 3.
Power results for Simulation I (scale-free structure) when causal nodes are hub nodes.
The power at α = 0.05 were based on 250 simulation replications for the interaction test and the conditional test of Module 1. The X-axis indicates the number of causal nodes out of the 20 nodes in a module. The three panels under each test, i.e., Low, Median, and High, indicate the level of the R2 explained by the module effects.
Fig 4.
Power results for simulation I (scale-free structure) when causal nodes are random nodes.
The power at α = 0.05 were based on 250 simulation replications for the interaction test and the conditional test of Module 1. The X-axis indicates the number of causal nodes out of the 20 nodes in a module. The three panels under each test, i.e., Low, Median, and High, indicate the level of the R2 explained by the module effects.
Fig 5.
Power results for simulation II (non-scale-free structure) when causal nodes are hub nodes.
The power at α = 0.05 were based on 250 simulation replications for the interaction test and the conditional tests. The X-axis indicates the number of causal nodes out of the 20 nodes in a module. The three panels under each test, i.e., Low, Median, and High, indicate the level of the R2 explained by the module effects.
Fig 6.
Power results for simulation II (non-scale-free structure) when causal nodes are random nodes.
The power at α = 0.05 were based on 250 simulation replications for the interaction test and the conditional tests. The X-axis indicates the number of causal nodes out of the 20 nodes in a module. The three panels under each test, i.e., Low, Median, and High, indicate the level of the R2 explained by the module effects.
Table 2.
Testing results from the baseline analysis of the Aspirin Data.
Table 3.
Testing results from the differential analysis of the Aspirin Data.