Abstract
Selection bias in sire''s estimated transmitting ability for milk yield in 2nd lactation due to prior selection on milk yield in 1st lactation were studied by a simulation technique, varying the selection intensities between 30 and 100% and applying 2 alternatives for carrying out the selection. In selection alternative I, cows with the highest milk yield for 1st lactation within a herd-year-season subclass were selected to give a 2nd-lactation record, while in alternative II cows were randomly selected. A total of 560,000 records were generated, assuming a heritability of 0.25 and 0.20 for milk yield during 1st and 2nd lactation, respectively, a genetic correlation of 0.90 and a repeatability of 0.40. Sires were assigned to 2 sire groups, corresponding to proven and unproven sires with a preassigned difference of 0.2857 .sigma.p1. Genetic values were estimated by single- and multiple-trait mixed model (MM) procedures and a combination of single-trait MM sire solutions using selection index theory. Selection bias within practicable selection intensities was shown to give quite small effects in single-trait MM estimates of transmitting abilities, as well as estimates of sire-group differences. At increased selection intensities the selection bias was quite obvious, when single-trait MM procedures were used. A correction of 2nd-lactation records for selection, followed by evaluation by single-trait MM, gave good results within sire group, but underestimated the differences between them. The combined index procedure using single-trait MM solutions also showed excellent results within sire group and is comparable to the multiple-trait MM procedure. The latter also gave unbiased estimates of differences between sire groups. Under Swedish conditions, single-trait MM procedures can be used for sire evaluation of 2nd-lactation milk yield, possibly combined with an index procedure to increase the accuracy of 2nd lactation sire evaluations.