Table 4 Summary of the models fitted to the full data set
ModelaFormb of umForm of ueSTY or METLog-LikelihoodAIC
PhenotypiccEDIAGDIAGdSTY−11,074.622,163.1
ECSCSeMET−10,877.021,768.1
ECS+DIAGCS+DIAGfMET−10,872.521,760.9
EFAM1, EUSFAM1g, UShMET−10,872.421,760.8
StandardiSDIAGDIAGSTY−10,924.921,863.8
SCSCSMET−10,794.021,602.1
SCS+DIAGCS+DIAGMET−10,790.621,597.2
SFAM1, SUSFAM1, USMET−10,787.821,591.6
PartitionedjPDIAGDIAGDIAGSTY−10,876.421,770.8
PCSCSCSMET−10,747.221,512.5
PCS+DIAGCS+DIAGCS+DIAGMET−10,744.621,511.2
PFAM1, PUSFAM1, USFAM1, USMET−10,744.221,510.4
  • STY, single trial year; MET, multi-environment trial; AIC, Akaike information criteria.

  • a All models derive from Equation 1 but are special cases of g(Equation 2).

  • b Details of forms of um are given in Table 1.

  • c Phenotypic model has g = ue.

  • d DIAG implies the covariance between the two trials is assumed to be zero, and is equivalent to fitting the two trials separately.

  • e CS is the compound symmetry model.

  • f CS+DIAG is the model described by Cullis et al. 1998.

  • g FAM1 is the factor analytic model (Smith et al. 2001), with main effect with k the number of factors equal to 1.

  • h US is the unstructured model (US), for two trials this model is equivalent to the FAM1 model.

  • i Standard RR-BLUP model has Embedded Imageum.

  • j Partitioned RR-BLUP model has Embedded Image.