Table 7 Average mean square errora (SD) of partitioned verses standard RR-BLUP model for cross-validation (Table 3), models (Table 4) and effects used to generate the GV and GEBV
ComparisonPhenotypic ModelLine Effects to Generate GVbStandard ModelPartitioned ModelMarker Effects to Generate GEBVcMean Square Error (SD)
Standard ModelPartitioned ModelStandard ModelPartitioned ModelStandard ModelPartitioned Model
5ECS+DIAGTotal 2010SCS+DIAGPCS+DIAGTotal 20108.01(0.75)7.74(0.70)7.85(1.09)7.53(1.04)7.73(1.55)7.35(1.49)
6ECS+DIAGTotal 2011SCS+DIAGPCS+DIAGTotal 20117.02(0.57)6.79(0.55)6.88(0.77)6.61(0.75)6.75(1.02)6.44(1.01)
7EFAM1Total 2010SFAM1PFAM1Total 20109.98(0.87)9.55(0.83)9.87(1.39)9.34(1.34)9.80(2.07)9.19(1.96)
8EFAM1Total 2011SFAM1PFAM1Total 20117.02(0.62)6.81(0.61)6.88(0.87)6.63(0.83)6.79(1.22)6.49(1.13)
  • GEBV, genomic estimated breeding value; GV, genotypic value; SD=standard deviation.

  • a The mean square error value is from a linear model for the validation set, in which the GEBV is the covariate and the GV the response. The mean square error shown is the average of the mean square error over the different number of iterations (Table 2). Lower mean square error indicates more accurate and precise estimates of GEBV.

  • b GV are calculated using a phenotypic model with all of the lines.

  • c Marker effects from DIAG form are in bold with year of trial shown, and are equivalent to results from a single trial year analysis, marker effects for the MET analyses are in bold and italic, three marker effects are possible: main, interaction 2010, interaction 2011, with the sum of the (main + interaction) marker effects being equivalent of a total marker effect for a particular year.