The traits were then treated as liabilities for binary diseases with prevalence 0.2. Disease status was simulated prospectively, as in a cohort study. The polygenic score was estimated and tested using both linear and logistic regression. Table S2 shows estimates of power and AUC compared to the analytic values. Results for the shrinkage estimator are identical to the regression estimator and are not shown. All simulations agree well with the analytic results, and the proposed estimating equations are accurate. The results for logistic regression agree well with those for linear regression, justifying the use of the latter to derive the analytic results.
