Vierhout C N, Cassell B G, Pearson R E
Department of Dairy Science, Virginia Polytechnic Institute and State University, Blacksburg 24061-0315, USA.
J Dairy Sci. 1998 Sep;81(9):2524-32. doi: 10.3168/jds.S0022-0302(98)70145-6.
Our purpose was to examine differences among AI progeny test programs in their effectiveness in identifying elite young sires for milk and protein. Data were from 6238 Holstein sires in five animal model evaluations of the USDA from January 1995 to February 1997. Bulls were required to be < or = 8 yr of age and to have > or = 10 daughters in production at the time of the evaluation. Bulls in AI programs were placed in nine groups based on affiliation with major AI organizations; from these nine groups, bulls with daughters averaging < 150 DIM and bulls sampled in > 100 herds were placed in two separate groups. Bulls sampled in organizations with controller numbers > or = 30 and bulls that had no connection to a sampling organization formed two additional groups. We also classified bulls by sampling codes of the National Association of Animal Breeders. A model predicting daughter yield deviation included effects of organization or sampling code, parent average (free of progeny information), and the interaction of organization or sampling method and parent average. When all data were used, a common intercept for milk was appropriate for all sampling methods, but the slopes differed (R2 = 0.44). Neither a common intercept nor a common slope was appropriate for protein. When data were restricted to the nine major organizations, a common intercept and slope were appropriate for milk, and R2 decreased to 0.14. A common intercept and slope were found for protein, and R2 decreased to 0.15. We detected no important differences in response to pedigree selection among progeny-testing methods used by major organizations that provide semen, but a difference was detected among the sampling codes of the National Association of Animal Breeders.
我们的目的是研究人工授精后代测试程序在识别优质年轻奶牛种公牛的产奶量和蛋白质含量方面的有效性差异。数据来自1995年1月至1997年2月美国农业部进行的五次动物模型评估中的6238头荷斯坦种公牛。公牛要求年龄≤8岁，且在评估时至少有10头女儿处于产奶期。根据与主要人工授精组织的隶属关系，将人工授精项目中的公牛分为九组；从这九组中，将女儿平均泌乳天数＜150天的公牛和在超过100个牛群中采样的公牛分为两个单独的组。在控制编号≥30的组织中采样的公牛以及与采样组织没有关联的公牛又形成了另外两组。我们还根据美国动物育种者协会的采样代码对公牛进行了分类。一个预测女儿产量偏差的模型包括组织或采样代码、亲本平均值（不包含后代信息）以及组织或采样方法与亲本平均值的交互作用。当使用所有数据时，对于所有采样方法，产奶量的共同截距是合适的，但斜率不同（R² = 0.44）。对于蛋白质含量，既没有合适的共同截距也没有合适的共同斜率。当数据仅限于九个主要组织时，产奶量有合适的共同截距和斜率，R²降至0.14。对于蛋白质含量也发现了共同截距和斜率，R²降至0.15。我们发现，提供精液的主要组织所使用的后代测试方法在对系谱选择的反应上没有重要差异，但在美国动物育种者协会的采样代码之间发现了差异。
