Fig. 8.

Prediction accuracy for the phi statistic (left), the proportionality coefficient using a clr transformation (centre) and the proportionality coefficient using the alr transformation with an unchanged gene (right). Each grey dot represents a gene pair. The applied cut-offs of 0.05 (for $\mathit{\varphi }$) and 0.98 (for $\mathit{\rho }$) are denoted by red vertical and horizontal lines, and the resulting quadrants of false positives (FP), false negatives (FN), true positives (TP) and true negatives (TN) are indicated. The range was chosen so as to include all false positives. The insets in the centre are Venn diagrams, where the left and right circles denote the set of pairs found on absolute data and on relative data, respectively. Shown are the numbers of FN, TP and FP. The circle in the intersection denotes the largest set of pairs having 100 % specificity that can be obtained by just adjusting the cut-off. Cut-offs for these sets are <0.012, >0.994, and >0.983, respectively (colour figure online)