EFFICIENCY OF ESTIMATION OF VARIANCE AND COVARIANCE COMPONENTS FROM FULL-SIB GROUP MEANS FOR CONTINUOUS OR BINARY RECORDS

High fecundity in fish allows for testing of large full-sib groups and testing of sub-samples of each group for different traits in differen t environments, and this benefit is especially useful when traits cann ot be measured on the breeding candidate itself. A MIVQUE-method that compresses the information in such data, by estimating variance compon ents based on full-sib mean data, Is presented for a two-trait model. The efficiency of this method is compared in a simulation study with a model using an individual observation vector when estimating additive genetic and residual (co)variances. The effect on efficiency of conti nuous versus binary traits, of varied average number of full-sibs, of varied frequencies in the binary traits, of different combinations of frequencies in the two binary traits, of different heritabilities, of different genetic correlations, and of two different design sizes were studied. Generally, all parameters influenced the efficiency, and thi s influence was larger on the residual, than on the additive generic v ariance component. The proposed methodology will be especially well-su ited for analysing binary tracts because the residual variance compone nt can be calculated via the phenotypic variance, which can be estimat ed on basis of the frequency of the trait.