Figure 6

Scatterplot of all pairwise correlations of the 1000 most variable genes in the yeast data. The blackest hexagons represent 7,708 pairs of genes. The lightest hexagons represent one pair of genes. Notice that, unlike the comparison of the biweight correlation vs. Pearson correlation, here there are fewer points in groups 2 or 3 (see section 2.1) indicating that the Spearman correlation is less able to recognize pairs that have mistakenly been given high Pearson (when resistant measure is opposite) or zero Pearson correlations (when resistant measure is high.) Also note that when there is less agreement overall between Spearman and Pearson than between biweight and Pearson (the center section is wider than in figure (1).)