Abstract
To argue for increased sample collection for disorders without significant findings, researchers retorted to plotting, for multiple traits, the number of significant findings as a function of the sample size. However, for polygenic traits, the prevalence of the disorder confounds the relationship between the number of significant findings and the sample size. To adjust the number of significant findings for prevalence, we develop a method that uses the expected noncentrality of the contrast between liabilities of cases and controls. We empirically find that, when compared to the sample size, this measure is a better predictor of number of significant findings. Even more, we show that the sample size effect on the number of signals is explained by the noncetrality measure. Finally, we provide an R script to estimate the required sample size (non-centrality) needed to yield a pre-specified number of significant findings.