Influence of population structure on estimates of direct and maternal parameters
The impact of population structure on the estimation of (co)variance components for traits affected by maternal effects was investigated through stochastic simulation. Two traits in a closed breeding herd of 400 cows and 50 sires with overlapping generations and random mating was simulated over nine generations. Simulations considered five different population structures on the basis of different proportions of dams with performance records (mis = 0, 0.1, 0.5, 0.8, and 0.9); three genetic correlations (r = -0.5, 0.0, +0.5) between direct and maternal effects; and three different values of genetic correlations (t = 0, 0.3, 0.8) between two traits. Three different parameter sets based on the ratio between direct and maternal genetic variances (par = 1:3, 1:1, and 3:1) for each trait of interest were also considered. Variance components were estimated by restricted maximum likelihood (REML) through DMU (a package for analyzing multivariate mixed models). The mis had no influence on SE of estimates of maternal, direct, and permanent environmental variances, but for direct-maternal covariance estimates, larger SE were obtained when mis was high (0.8 or 0.9) and r was negative. The influence of the parameter sets on the SE of the estimates was greater than that of the population structure. The t had no major influence on the SE of the estimates. The mis and r had greater effects on bias of estimates than the par and t. Bias was obtained when mis was high and r was positive. The largest biases were detected when the direct variance was greater than the maternal variance, as would be the situation of most growth traits in livestock, particularly in beef cattle. Total bias (%) of the whole set of parameters were large when mis was 0.9, r was positive, and the variance ratio between direct and maternal was equal (1:1). In general, results of this study showed that the par and r had more influence on the SE of the (co)variance estimates than mis and t. However, mis and r had greater effect on bias of the estimates. Complete data sets with more links of dam-offspring are required for accurate estimation of maternal genetic parameters in the single and multiple trait situations.