Figure 2
Relationship between the significance levels of different set test statistics. Scatter plots of all pairwise combinations of significance between the set test statistics, i.e. TCVAT (CVAT), TScore (Score), TSum using single marker effects \(({\sum \widehat{{\rm{s}}}}^{2})\) or using single marker t-statistics \((\sum {{\rm{t}}}_{\widehat{{\rm{s}}}}^{2})\), and TCount with a threshold p-value < 0.05 \(({\rm{I}}\{{\rm{\Pr }}({{\rm{t}}}_{\hat{{\rm{s}}}}) < 0.05\})\) and p-value < 0.01 \(({\rm{I}}\{{\rm{\Pr }}({{\rm{t}}}_{\hat{{\rm{s}}}}) < 0.01\})\). Significance, shown as −log(p), was measured for the association of simulated phenotype with genomic feature over a dilution range of adding 0 to 2000 non-causal SNPs to the C 1 causal set. Plots are arranged such that all plots in a row share a common y-axis, and all plots in a column share a common x-axis. The names of the x- and y-axes are shown in the diagonal boxes. Genomic heritability was set to 30% (h2 = 0.3), and the proportion of genomic variance explained by the feature was 30% (\({{\rm{h}}}_{{\rm{f}}}^{2}\) = 0.3). The random causal model was used, randomly selecting causal SNPs (C 1 and C 2) from the complete set of SNPs. Five replicates were used within each line (Nrep = 5).
