Figure 2.
PRS-FHlog and PRS-FHliab increase prediction accuracy in simulations
We report mean liability-scale R2 across 10 simulations for PRS alone, FH alone (FHlog and FHliab), and PRS-FH methods (PRS-FHlog and PRS-FHliab) for different values of disease prevalence. Error bars denote standard errors. Numerical results are reported in Table S2.