Figure 7

Simulation Results. A simulation demonstrating the distribution of the three p-value derived parameters: μ or average rank (x-axis), SAD (y-axis) and ε or errval (z-axis) for 100 sites. The sites are generated based on four replicate datasets with significant reproducibility (Spearman's ρ = 0.93). The seeding is based on the p-value distributions. The ideal sites with maximal rank consistency are represented by the vertex cluster at [μ, SAD, ε] = [0,0,0]. Scatter is introduced in the dataset via artificial introduction of noise. This causes gradual migration of sites from the vertex cluster to the top right hand edge of the graph, where the sites represent the worst intra-replicate ranks, least inter-replicate rank consistency and lowest statistical significance. Despite the introduction of significant level of noise, the density of the vertex cluster remains the highest.