MONTE-CARLO STUDY OF (CO)VARIANCE COMPONENT ESTIMATION (REML) FOR TRAITS WITH DIFFERENT DESIGN MATRICES

Based on simulated data sets, covariance matrices were estimated using REML in a multivariate mixed linear model. Simulated traits were: dai ly gain and backfat in the field and daily gain at the station. Thus, the design matrices are completely different for field and test statio n traits. in 4 variants the proportion of the station test animals was systematically modified (2 female offspring each from 10%, 30%, 50% a nd 70% of litters). Comparisons are based on the classical statistics bias, mean square error as well as the ratio of achieved (R) and true response (It). This ratio exhibit clear differences for the variants c ompared (greatest loss on selection gain of 35% in variant 1, 5% in va riant 4).