Smith E J, Savage T F
Department of Animal Sciences, Oregon State University, Corvallis 97331-3402.
Poult Sci. 1992 Feb;71(2):229-34. doi: 10.3382/ps.0710229.
A comparison of four methods for variance component estimation to compute heritability of embryonic mortality in the early (Days 1 to 10) and late (Days 21 to 28) incubation period, was performed using Henderson's Method 3, maximum likelihood (ML), restricted maximum likelihood (REML), and minimum variance quadratic unbiased estimator (MIVQUE). Incubation records from Wrolstad Medium White turkeys divergently selected for semen ejaculate volume were the source of data for the present study. Negative sire components of variance were obtained for early mortality by both Henderson's Method 3 and the MIVQUE procedure. Estimates of the sire variance component for the late mortality period were lower (P less than .05) for ML than those obtained using Henderson's Method 3, REML, and MIVQUE. It was concluded that likelihood procedures (ML and REML) are superior for estimating variance components and genetic parameters from unbalanced data.
采用亨德森方法3、最大似然法（ML）、限制最大似然法（REML）和最小方差二次无偏估计法（MIVQUE），对四种方差成分估计方法进行比较，以计算早期（第1天至第10天）和晚期（第21天至第28天）孵化期胚胎死亡率的遗传力。本研究的数据来源于对Wrolstad中白火鸡精液射精量进行差异选择的孵化记录。通过亨德森方法3和MIVQUE程序均获得了早期死亡率的负父系方差成分。晚期死亡率阶段的父系方差成分估计值，ML法低于（P<0.05）使用亨德森方法3、REML法和MIVQUE法获得的估计值。研究得出结论，似然法（ML和REML）在从不平衡数据中估计方差成分和遗传参数方面更具优势。
