The use of mixed models to analyse experiments in animal production was discussed taking into account that, in general, the animals used in the experiments are relatives (in some cases, close relatives as full or half sibs), An exatnpk to estimate the effects of year-season, parity order and some covariates on litter size and weight traits and kindling interval was given using three sets of genetic parameters that fall within the range of parameters reported in the literature. Results obtained from such example show that it is not necessary to have very accurate values of the genetic parameters to be used in the mixed model to get the estimates of the fixed effects. This concept was proved in results of the present study since the estimates of least square means of the fixed factors (year-season, parity order) and the regression coefficients of the covariates remain unchanged with the change of the parameter set used to solve the mixed model equations. Also, results of the tests of significance are always the same for the three sets of genetic parameters. In conclusion, to get inferences about fixed effects, the mixed model methodology seems to be robust to non-negligible changes in the parameters used. |
