Table 4 Mean and SE of the prediction accuracy values for the parametric and the nonparametric methods for the F2 population with heritability h2 = 0.70
F2, h2 = 0.70, AccuracyAdditive MeanEpistatic MeanAdditive SEEpistatic SE
Least squares regression0.560.090.050.06
Ridge regression0.800.020.020.07
Bayesian ridge regression0.800.010.020.07
BLUP0.800.010.020.08
LASSO0.82−0.010.020.05
Bayes LASSO0.810.010.020.07
Bayes A0.810.000.020.07
Bayes B0.810.010.020.07
Bayes C0.810.010.020.07
Bayes Cπ0.830.010.020.07
Nadaraya-Watson estimator0.670.350.040.06
RKHS0.760.290.030.05
Support vector machine0.780.330.030.07
Neural network0.770.050.030.09
  • Mean and SE of the prediction accuracy values for both the additive and the epistatic cases. The first 10 methods are parametric and the last four are nonparametric. The calculations for the epistatic mean and epistatic SE for the LASSO method are based on 213 replicates, for the epistatic mean and epistatic SE for the neural network method they are based on 493 replicates, and, for the rest, the calculations are based on 500 replicates.