Boxplot showing the within data set cross-validation consistency.

<p>For the 30 data sets (a) enrichment scores and (b) average B-H adjusted <i>p</i>-values of enrichment scores. Each data set was randomly halved. Spearman correlation of genes' aCGH and expression values was used to rank genes in each half data set. The top 10 from the first half was used as a gene-set and scored for enrichment in the second half. This was repeated for 10 random divisions of each data set.