Relative power vs. number of associated traits for different global testing methods.

Relative power at significance level 0.001, based on 10³ simulated replicates, is plotted against the number of associated traits out of 10⁴ total traits tested, for 7 of the 8 global testing methods considered. The curve for ARCHIE is visually indistinguishable from that for Sum-χ², so it is not plotted separately. For a given number of associated traits, relative power for a method is defined as its power divided by the maximum power achieved by any of the 8 methods for that setting. For each point of the plot, the corresponding vertical bar represents the 95% confidence interval.