F. Canavesi et al., SIRE-BY-HERD INTERACTION EFFECT WHEN VARIANCES ACROSS HERDS ARE HETEROGENEOUS .2. WITHIN-HERD VARIANCE-COMPONENT ESTIMATES, Journal of animal breeding and genetics, 112(2), 1995, pp. 107-116
Six different variance structures were simulated across herds. Data st
ructure corresponded to a sample data set and comprised 197 335 lactat
ions of 82 062 cows distributed in 924 herds. The homogeneity case, as
well as different combinations of heterogeneous variances, were consi
dered. Variances were varied assuming a positive correlation of 0.4 wi
th herd-year-season (HYS) average production. Two animal models, with
and without sire-by-herd (S x H) interaction, were compared through wi
thin-HYS variance components estimates. A regression-analysis method w
as used for the esrimarion, Correlations between estimated and true va
riance components within HYS were around 0.7 for subclasses with al le
ast 20 observations. The model fitting a S x H interaction showed a re
duced correlation with true values and HYS average production, but the
reduction, although significant, was very small (-0.02). Thus, the ef
fect of the interaction in absorbing heterogeneous variance was very s
mall.