Fig. 4.

Training accuracy of two models measured by Pearson’s correlation $(Cor{r}_{\widehat{\mathbf{y}},\mathbf{y}}^{(P)})$ between observed and fitted phenotype under different simulation settings. ADD additive model, POE parent-of-origin effects model. $s=$ proportion of imprinted QTL; $\mathit{\rho }=0$ and $\mathit{\rho }=1$ denote complete imprinting and no imprinting, respectively