Figure 5.

Reproducibility of normalization methods for different degrees of spike-skew in simulated data. The SD of the random multiplicative spike-skew term in the simulations was adjusted from 0.1 to 0.4 (10-40%). Increasing spike-skew specifically degrades the performance of the F metric. Note that the relatively poor performance of F relative to F^s and AD^s when the spike-skew is 0.2 (20%) is similar to that observed in the experimental data (Figure 3). Twenty simulated hybridizations were generated for each level of spike-skew.