Type 1 error rates using pathways with 15% missing edges.

Error rates were calculated from score tests on 1000 simulated data sets using graphs with 15, 30, or 45 nodes. The graph used to simulate Z and Y was of high edge density, while the graph used to test was of low density. The low density graphs are drawn from the Barabasi-Albert model with edge density 0.13, 0.07, and 0.04 for graphs with 15, 30, and 45 nodes, respectively. High edge density graphs are created by giving any 2 nodes without a direct edge between them a 15% chance of becoming directly connected. This creates graphs with an average edge density of 0.26, 0.21, and 0.19 for graphs with 15, 30, and 45 nodes, respectively. Pathway information was included in kernel score test through the normalized Laplacian .
(XLSX)