MONTE-CARLO SIMULATION IN VARIANCE-COVARIANCE-COMPONENT ESTIMATION INMIXED LINEAR MULTIPLE-TRAIT MODELS

The efficiency of mixed-model techniques for the estimation of varianc e-covariance-matrices (VCM) was investigated by a Monte-Carlo study (o ver 12 generations with 30 replications) in a multi trait model. Three methods for estimating the VCM (ANOVA with correction of the fixed ef fects via OLS, HENDERSON III, REML) were compared through their biases , MSE and the realised selection response in the 12 generation. ANOVAK OR and HENDERSON III leads to an important increase in bias and MSE in relation to REML. The selection response compared with ANOVAKOR was 1 04% and 106% for HENDERSON III and REML resp. in the aggregate genotyp e (egg number + 3 feed conversion).