Estimation Of Genetic Parameters And Covariance Functions Of Body Weight In Corriedale Sheep

Data set of 3332 body weight records at birth, weaning, and 8, 12, 18 months of age from1269 Corriedale sheep was used to estimate genetic parameters and covariance functions. TheGLM procedure of SPSS software was used to select non-genetic factors as fixed effectswhich genetic analysis models included. In the genetic analysis models, non-genetic factors were main effects of birth year, birth month, birth date, birth type and sex and all 2-way interactions among them. Firstly, parameters under 9 models including different random effects were estimated by unitrait analysis, and Likelihood Ratio Test (LRT) was used to select the most suitable model; Secondly, parameters under the most suitable model were estimated by multi-trait analysis; Thirdly, covariance functions were estimated by random regression model under the most suitable model, of which random effects were fitted from 3 to 5 orders, and LRT, Araike's Information criterion (AIC) and Bayesian Information criterion (BIC) were used to select the most suitable order of fit. Independent variables were Legendre polynomials of age at recording. Mean trends were modeled through a quartic regression on orthogonal polynomials of age. Five different measure errors were fitted for weights at 5 different ages. The variance components from unitrait analysis and the covariance components from multi-trait analysis were used as the starting values of estimations of covariance function. In addition, when fixed effects were not selected, or covariances among errors were or not considered, changes of parameter estimates were studied in this paper. The results were as follows:1 Birth year and birth type had significant influence on birth weight; sex, birth year, birth month and birth type all had significant influence on weights at weaning, 8 months, yearling and 18 months.2 For unitrait analysis, when fixed effects were or not selected, the most suitable models both were model 3: y = Xb + Z₁a + W₂c + e, and the difference between the estimates of the corresponding genetic parameters was not significant, but the equation number involved was fewer after GLM's significance test, the efficiency of computation was much improved, and the estimates of the corresponding e were smaller under most models.3 For multi-trait analysis, when considering the covariances among errors, the estimates of genetic parameters agreed with but were slightly better than those of unitrait analysis. Heritability estimates were 0.28, 0.33, 0.30, 0.23 and 0.21, for weights of birth, weaning, 8, 12 and 18 months, respectively. The estimates of genetic correlation were in the range from 0.37⁰.96. When not considering the covariances among errors, the estimates of genetic parameters were higher.4 For estimation of covariance function, the estimates of genetic parameters were consistent with the results from multi-trait analysis not considering the covariances among errors.5, When the different orders were fitted, the most suitable orders of fit for direct additive genetic effect and maternal permanent environment effect were 5 and 4, respectively.6 Compared with full order of fit, reduced fit had less parameters, improved computing efficiency, reduced sampling error. It thus could be better in modeling the genetic variation of body weights growth, and had a suitable covariance structure.7, The estimated covariance function for direct additive genetic effect was...