Estimation of variance components with the finite polygenic model (FPM) was evaluated.
Phenotypic data for a 6300-pedigree simulated under a wide range of additive genetic models were
analysed with constant homozygote difference across loci using deterministic Maximum Likelihood
(DML) and a Bayesian method implemented via Gibbs sampling (BGS). Results indicate that
under no selection, both DML and BGS accurately estimated the variance components, with a
FPM of 5 loci or more. When both analysis methods were applied to equivalent data sets on
populations that had undergone selection, the DML method produced upward biased estimates of
additive genetic variation and heritability due to its use of pedigree loop cutting, while BGS
provided more accurate estimation. BGS was extended to non-additive FPMs with variable
homozygote differences and dominance effect across loci. This method was used to analyse data
simulated under two genetic models with positive, completely dominant gene action at all loci.
Results indicate that the estimates of additive and dominance variances slowly increase as the
number of loci in the FPM for analysis increases, while accuracy of predicting individual breeding
values and dominance deviations remains unaffected. For the simulated pedigree structure, a FPM
with 10 loci or slightly fewer appears to be appropriate for variance component estimation in the
presence of dominance.