Figure 1.

Estimated loadings (after varimax rotation) of each of the four environments on the first and second common factors of G0P obtained from models using a pedigree-based (PFA-UN) and marker-based (MFA-UN) additive relationship matrix. The xy coordinates are the estimated loadings after varimax rotation, and the numbers in the body of the plot are the environment labels (1–4). In both models, G0 was modeled using a two-common factor model, and the residual covariance matrix, R0, was completely unstructured.

 


Figure 2.

Average (across four environments) correlations between predicted and observed performance derived from models using only pedigree (Model 9 PFA-UN), only markers (Model 18 MFA-UN), or pedigree + markers (Model 27 PMFA-UN) for two cross-validation schemes (CV1 and CV2). The right vertical axis shows the probabilities of one rank change for the given correlations.

 


Figure 3.

Correlation between predicted and observed performance in Environment 1 (E1) and average of Environments 2, 3, and 4 (E2, E3, and E4) obtained in CV2 using (a) pedigree, (b) markers, and (c) pedigree and marker–based models with different specifications for the residual and genetic covariance matrices (D = Diagonal, FA = two-common factors model, UN = unstructured). The right vertical axis shows the probabilities of one rank change for the given correlations.

 


Figure B1.

Relationship between Pearson's correlation coefficient and the probability that there will be a rank change between any two given genotypes if the observed or the predicted genotypic value is used.