Comparison of Heritability Estimates from Daughter-Dam Regression and Paternal Half-Sib Correlation

Authors:

Abstract
Analysis of 60,000 pairs of daughter and dam first-lactation records ex- pressed as deviations from herd-mate averages for five breeds yielded markedly different heritability estimates from daughter-dam regression than from paternal half-sib correlation. These results suggest that 18% of the within-herd variation is due to genetic maternal effects. The same records analyzed as mature equiva- lent records and not as deviations but by a sire by herd model did not show this difference. Confounding between year and sire effects probably biased the intra- class correlation estimates upward in the latter analysis, since year effects were not included in the statistical model. Heritability estimates of milk production are numerous. Lasley (3) lists an average based on 30 such reports. Comparisons of estimates from daughter-dam regression and paternal half-sib correlation, however, are not frequent. Bradford and Van Vleck (1) reported a value of .44 from regression for 2,580 pairs and .25 from the intrasire correlation for first lactation de- viations from herd-mate averages for the same artificially sired Holstein cows. Rendel et al. (4) and Gravert (2) both reported better agreement between the two methods of estimation used on the same data. Van Vleek and Bradford (5) later found almost identical estimates of heri- tability from daughter-dam (.448) and grand- daughter-granddam (.444) regression on 10,419 first-lactation trios. The corresponding paternal half-sib estimate was .23. The estimated vari- ance components from a sire by herd model for mature equivalent records and deviations gave heritability estimates of about .40 for ma- ture equivalent records from intraherd daugh- ter-dam and granddaughter-granddam regres- sion and also from paternal half-sib correla- tions. The analyses reported here are for larger numbers of daughter-dam pairs for the major dairy breeds. The purposes of the analyses were to determine whether the discrepancy be- tween estimates was true for all breeds and to confirm that analyses of deviations and mature equivalent records give different results. DATA