Estimation of Variance Components in Populations Selected over Multiple Generations

Publisher Website
Google Scholar
Abstract
Methods for estimating variance components in selected data were compared using simulated data from a selection experiment. After five generations REML-method using full relationship matrix and recursive prediction which take account for decreased additive variance due to selection gave unbiased estimates. REML-estimates had smallest variance, but the difference between the variance of REML and recursive prediction estimates was small. Realized heritability estimates were biased downwards and their variance was considerably larger. The design of the experiment had a clear effect on the variance of the estimates. The bias and the variance of the estimates increased when family selection was used instead of mass selection.