The Choice of an Appropriate Model for Evaluating Beef Sires by Best Linear Unbiased Prediction Using Field Data

Abstract

To find the most appropriate model for sire evaluation by the best linear unbiased prediction (BLUP), an investigation was done using the 2, 208 records of fattened Japanese Black steers which were shipped from Oita prefecture to MATSUBARA Carcass Market during 1980 to 1983. The least squares analysis of variance was done by the program LSMLMW written by HARVEY (1985). Twenty three sires which had more than 10 progeny were evaluated by the BLUP of a semi-reduced animal model in which the evaluated animal and the recorded animal were completely separated. In terms of residual variance and predicted error variance (PEV), an appropriate model was chosen. A model including maternal grandsire without grouping of sires appeared to be superior to models which include grouping of sires. Fattening farms on a small scale had better be excluded, while all sires should be included even if they had one progeny.