Fig. 9
From: Priors, population sizes, and power in genome-wide hypothesis tests
Non-oracular priors for GWAS. Non-oracular priors, with accuracy f defined as the fraction of true positives included in the tested set, can perform worse that not using a prior at all. Equal performance (solid line) of a non-oracular prior and of no prior is defined by the \(R^2\) that can be detected with 50% power (left), 80% power (middle), or 95% power (right) at \(p = 5\times 10^{-8}\) typically used for GWAS. Performance is depicted for non-oracular priors having accuracy f from 0% to 100% (equivalent to an oracular prior) and having prior strength S from 1 to 1000. Accuracies for equal performance at prior strength 10, 100, and 1000 are shown (labeled points). Above the line, the non-oracular prior performs better; below the line, the non-oracular prior performs worse and should be avoided