Author
 |
VAN VLECK L DALE - 5438-01-30 |
|
Submitted to: Journal of Animal Science Supplement
Publication Type: Abstract Only Publication Acceptance Date: 3/10/1995 Publication Date: N/A Citation: N/A Interpretive Summary: Technical Abstract: An option in the MTDFNRM program of the MTDFREML package is to use Westell's rules to calculate the inverse of the numerator relationship matrix augmented to correspond to selected groups (breeds) assigned as proxy parents for foundation animals. The resulting mixed model equations correspond to the Quaas-Pollak transformation to predict each breeding value directly as the sum of a predicted random deviation and the appropri ate function of fixed pure breed effects. An equivalent model is to in- clude as covariates, expected fractions of genes inherited from each pure breed. With either model, effects due to direct and maternal heterosis may need to be included. With Westell's rules and a maternal effects model, a hidden but implicit assumption is that all breeds be represented at least fractionally in maternal effects. The complexity of the resulting mixed model equations and the computing algorithm in MTDFRUN result in solutions for maternal breed effects that are not estimable even as differences from maternal effects of other breeds. The solutions may or may not appear to be reasonable. The only way to know for sure if differences in the mater- nal breed effects are estimable is to obtain expected values of the solut- ions. However, the expected value option in MTDFRUN does not work with the Q-P-W mixed model equations. In general, expected values for the group solutions from the Q-P-W equations are difficult to obtain. Expected values for solutions with the equivalent covariate model can be obtained but the need to use the covariate model negates the advantage of Westell's rules which automatically combine foundation breed effects in the proper proportions to predict the breeding value for each animal. |
