Effects of selection of data on estimates of breeding values by three methods for litter size, backfat, and average daily gain in swine.

Estimated breeding value was calculated based on individual phenotype (SP), an index of individual phenotype and full- and half-sib family averages (SI), or Best Linear Unbiased Prediction (BLUP). Traits considered were litter size (LS), backfat (BF), and ADG. Estimated breeding values were calculated using all data and after deletion of the poorest 5, 10, 15, or 20% of the records for BF and ADG, or 4.8, 8, 13, or 21% of the records for LS. When all data were used, expected genetic gain from BLUP was greater than for SP by 22, 7, and 31% and greater than for SI by 10, 4, and 21% for LS, BF, and ADG, respectively. Expected genetic gain was 4, 0, and 3% lower for LS, BF, and ADG, respectively, for selection on breeding values estimated by SI after the poorest 20% of the records were deleted compared with selection on estimates by SI using all the data. Genetic gain using BLUP on data with the poorest 20% of the records deleted was reduced by 5, 2, and 8% for LS, BF, and ADG, respectively, compared with genetic gain using BLUP on all the data. The advantage in genetic gain of BLUP, with 20% of the poorest records deleted, over SP was 15, 5, and 21% for LS, BF, and ADG, respectively. Although BLUP is affected to a greater degree by deletion of records than is SP or SI, selection of swine using BLUP on field data would improve response to selection over the use of SP or SI.