Fig. 2

Comparison of EmProc, CorrBinom, IndJZ, and McNemar in terms of hypothesis testing and pointwise confidence intervals to compare hit enrichment curves for competing algorithms. (A-B): Estimated power of the hypothesis test to detect differences between two competing algorithms, where each algorithm follows a bibeta model and scores are correlated with \(\rho =.1\) (A) or \(\rho =.9\) (B). C, D: Estimated coverage probability (C) and average width (D) of pointwise confidence intervals for the difference in hit enrichment curves for two competing algorithms, where each algorithm follows a binormal model and scores are correlated with \(\rho =.9\). Simulations were conducted with 10,000 Monte Carlo replicates. Shading show the Monte Carlo estimate ± 1.96 times the Monte Carlo standard error