Fig. 1

Empirical power of LASSO, EN, RR and LMM based on simulations. X-axis represents the proportion of variance explained by the simulated effect (range from 0.01 to 0.50). 1SE: in this model, the tuning parameters were determined based on the one standard error rule; Min: the tuning parameters were determined by minimizing cross-validated mean squared prediction error; P: the p-value threshold was used for statistical significance, which was set at 0.05/15,092 for LMM, and 0.05 for LASSO, EN, and RR. LASSO least absolute shrinkage and selection operator, EN elastic net, RR ridge regression