Summary
Deterministic methods were used to calculate the accuracy of evaluation, and the correlation of estimated breeding values (EBV) among sibs, when evaluation is based on phenotypic records of the individual and relatives, or on quantitative-trait-locus (QTL) information. An additive infinitesimal genetic model was assumed, with discrete generations of truncation selection on EBV, and random mating of selected parents in a hierarchical design. It was shown that the overlap of phenotypic data between the evaluations of relatives causes large increases in correlations of EBV with only modest increases in accuracy. This explains the disproportionate increases in inbreeding observed by other authors. By contrast, the use of major-locus information leads to low correlations of EBV among relatives, provided that functional loci are directly identified or there is complete linkage disequilibrium. Use of information from QTL should reduce EBV correlations and, hence, inbreeding, but use of information from markers may not be useful, since errors of estimation of association with QTL will be shared by family members, leading to increased correlations of EBV.