We simulated survival data under both the cohort and case-cohort study designs. HR was used as the underlying measure of association and is assumed to be the reference (or gold) standard, and as such, bias was defined as the estimated log(HR) or log(OR) minus the underlying log(HR) (i.e. . Precision as estimated by the model was assessed using the standard error (SE) of the estimated log(HR) or log(OR). Statistical power was defined (at the per SNP-test level) as the percentage of simulation replications where the SNP–disease association had a P-value less than the type I error rate of α=0.05.
