TABLE 4

Pearson correlations between predicted and actual values of the progeny average of each sire for late mortality in each subset (20% sires predicted in each subset) and by method

SubsetE-BLUPEmbedded Image-metricKernelRKHSBR
First0.030.260.050.27a0.13
Second0.180.250.280.370.12
Third0.18−0.130.06−0.010.17
Fourth−0.040.120.130.280.15
Fifth0.170.060.230.150.25
Global0.100.080.140.200.16
  • E-BLUP, Bayesian linear model without genomic information; Embedded Image-metric, linear regression on SNPs based on the Embedded Image-metric model; kernel, nonparametric kernel regression with SNPs within sire treated as a genomic combination; RKHS, reproducing kernel Hilbert spaces regression with SNPs within sire treated as a genomic combination; BR, Bayesian regression on 1000 SNPs based on Xu (2003).

  • a Higher values indicate more accurate predictions. The highest correlation for each set is in italics.