General Flexibility of Linear Model Techniques for Sire Evaluation

Abstract

Linear model methods applied to the mixed model provide a flexible and powerful tool for sire evaluation under a wide variety of situations. These techniques combine the known, desirable properties of selection index and the capability of linear model methods to deal with large sets of data with unequal subclass numbers. The problem to be solved is to find the best evaluation of a sire that is regarded as a random individual from some specified subpopulation (group). Defining as best that evaluation (prediction) which, in the elass of linear fimctions of the observations, is unbiased and has the smallest possible variance of prediction errors, the problem has been solved provided relative values of elements of the variance-covariance matrix of random variables in the model are known. The method derived to meet these specifications is Best Linear Unbiased Prediction. The computing technique is a simple modification of least squares.