Abstract
Monte Carlo computer simulation of a closed strain herd of swine was used to investigate the accuracy of genetic parameter estimates for litter size at birth by REML. A breeding herd of 10 sires and 50 dams in a base population was assumed; one male was randomly mated to five females. From each litter, one boar and one, two, or three gilts as three schemes were reared and one boar per five gilts was randomly selected and mated. Eleven generations including the base population, without overlapping, was simulated. The heritability of litter size at birth was assumed to be 0.1. Generation constants were generated as a fixed effect. A hundred replications were simulated for each scheme. For each generation of each replicate, means of the genetic parameter estimates and the square roots of their mean square errors (SMSE) were calculated. When the number of animals with a record of litter size at birth was less than 500, additive genetic variance and heritability for litter size at birth were overestimated and the environmental variance was underestimated. SMSE of each genetic parameter decreased rapidly with increase in population size, but it declined slightly when the number of animals with a record was more than 500. When the number of growing females per litter was one, heritability estimates of about 30% were less than 0.1% even if records of over five generations were used for genetic parameter estimates.
