Abstract
Predicted breeding values (PBV) are usually required only for potential parents, i.e., live animals. However, data from all animals (dead or alive) should be used in calculating the PBV of live animals to avoid certain types of selection bias and to improve accuracy of evaluation. Over time, a substantial data base may be collected, necessitating extensive computing effort. A method of combining a reduced animal model with recursive prediction is described that reduces the number of equations to be solved simultaneously. A numerical example is given and applications to within herd evaluation systems are discussed. Copyright © 1984. American Society of Animal Science . Copyright 1984 by American Society of Animal Science