Comparison of methods of estimating variance components in pigs

Components of variance due to average effects of genes (sigma 2g), environmental effects common to littermates (sigma 2c), and environmental effects peculiar to individual pigs (sigma 2e) were estimated (--) by the Pseudo Expectation Approach (PE). Data were litter size (LS), backfat (BF; centimeter) and ADG (kilograms/day) collected from the Nebraska Gene Pool swine population between 1967 and 1986. Mean square errors (MSE) for h--2 and c--2 (sigma--2g and sigma--2c divided by phenotypic variance) by PE and nested ANOVA and h2 estimated by offspring on parent regression (REGOP) were evaluated using simulation of 200 repetitions of the Nebraska Gene Pool population. Parameter values for sigma 2g, sigma 2c, and sigma 2e used in simulations were PE estimates from the Gene Pool population. Estimates of h2 from PE were .18 +/- .06 for LS, .56 +/- .06 for BF, and .16 +/- .05 for ADG. Estimates of c2 from PE were .01 +/- .03 for LS, .09 +/- .02 for BF, and .19 +/- .03 for ADG. Compared with REGOP, PE yielded h--2 with smaller MSE for BF and ADG and larger MSE for LS. The MSE of PE was smaller than the MSE of the nested ANOVA estimate for all estimates and traits. These results were interpreted to suggest that considerable gains in precision in estimation of genetic parameters could be achieved by accounting for all relationships in lieu of accounting for only half- and full-sib relationships or parent-offspring relationships.