INFLUENCE OF LONG-TERM SELECTION AND EFFECTIVE POPULATION-SIZE IN VARIANCE COMPONENT ESTIMATION - COMPARISION OF 3 MODELS WITH SIMULATED DATA

For a concrete data set the variance components can not be estimated w ith an animal model, because the individual phenotypic values can not be assigned uniquely to the animals within a full-sib group. The sire dam model and the random animal model are two potential models for var iance component estimation. At the random animal model the measured ph enotypic values are assigned randomly to the animals within a full-sib group. Then the variance components are estimated by an animal model. On simulated data sets the influence of long term selection and effec tive population size on the estimations with different models are inve stigated. The animal model is used for comparison. The variance compon ent estimation of this model (means over 50 replications, Ne=20,40,80) is independent of the population size and the number of generations. For a small population size (Ne less than or equal to 80) and long ter m selection a good conformity is shown between random animal model and animal model, where the difference increases when effective populatio n size increases and the difference decreases when the number of gener ations increases. The random animal model shows a better conformity wh en selection is going on over 20 and more generations for all variance components with an animal model than sire dam model does. These advan tages will become smaller or gets lost when effective population size increases for all variance components except environmental variance.