We showed previously that relationships among the parents can be modeled by including
pedigree information, which results in the GCA variance being properly adjusted to reflect the variance among unrelated samples of the reference population. ASReml internally makes this adjustment by using the numerator relationship matrix to model the additive effects of the parents. However, ASReml does not automatically adjust the SCA variance to account for the possibility that some progenies that appear to be half-sibs based on the immediate generation pedigree may in fact have a closer relationship and share dominance effects. Returning to the pine diallel example, progeny from the cross of parent 590 × 634 are closer than half-sibs to progenies from 626 × 634 (2θ = 0.3125) because parents 590 and 626 are both progenies of ancestor 333. Furthermore, progenies within the cross of 590 × 626 are more closely related than full-sibs, their additive relationship is 2θ = 0.625. ASReml accounts for the increased additive
relationships in these cases when it uses the pedigree information on the parents. It does not,however, account for the small covariance between the dominance effects of families 590 × 624 and 626 × 624 (Δ = 0.0625), or the slightly increased dominance covariance between progenies within the 590 × 626 family (Δ = 0.28125). Ignoring these relationships leads to biased estimates of the dominance variance based on the diallel SCA variance estimate. In the example shown, the bias is expected to be small since the changes in affected coefficients are small.