Figure 4. Optimal Design as a Function of Minor Allele Frequency and Effect Size.

Percentage of participants sequenced (x-axis) and total sample size (y-axis) under optimal designs to attain statistical power ≥80% for rare and common variants across two effect size values for each of the four study populations using the Infinium Core array. Here, effect size refers to the χ-squared non-centrality parameter (NCP) for single-variant association tests given perfect genotype accuracy, which is defined as η² in Methods. Relative risk (RR) values corresponding to each combination of MAF and NCP are indicated in the far-right panel (for Sardinians). With NCP held constant, differences in optimal design for different MAF values are solely due to differences in imputation coverage and quality across the MAF spectrum.