Table 2 Predictive abilities in computer simulations using LASSO, the elastic net, and RR-BLUP
Model complexity level ρ = p0/n
n/pMethod0.100.250.501.00
0.10LASSO0.69 ± 0.020.41 ± 0.020.20 ± 0.050.09 ± 0.04
Elastic net0.62 ± 0.020.35 ± 0.030.19 ± 0.030.09 ± 0.03
RR-BLUP0.20 ± 0.030.21 ± 0.030.22 ± 0.030.22 ± 0.03
0.25LASSO0.73 ± 0.010.46 ± 0.020.29 ± 0.030.13 ± 0.05
Elastic net0.73 ± 0.010.48 ± 0.030.33 ± 0.020.19 ± 0.02
RR-BLUP0.38 ± 0.010.35 ± 0.010.34 ± 0.020.34 ± 0.01
0.50LASSO0.77 ± 0.010.57 ± 0.020.39 ± 0.020.26 ± 0.02
Elastic net0.76 ± 0.010.60 ± 0.010.43 ± 0.010.31 ± 0.01
RR-BLUP0.47 ± 0.010.48 ± 0.010.49 ± 0.010.48 ± 0.01
1.00LASSO0.79 ± 0.0030.68 ± 0.010.52 ± 0.010.42 ± 0.02
Elastic net0.80 ± 0.0040.67 ± 0.0040.54 ± 0.010.45 ± 0.01
RR-BLUP0.61 ± 0.010.62 ± 0.010.61 ± 0.010.60 ± 0.01
  • Average predictive abilities ± SEs were estimated using fivefold cross-validation with 10 replications for each scenario. All scenarios were simulated according to procedure 1 (Figure S6), using p = 2000 independent markers and h2 = 0.75.