Correcting for Early Within-family Pre-selection in Genetic Evaluation - A Simulation on Rainbow Trout Growth

Abstract Text: Rainbow trout breeding scheme was simulated to estimate how within-family pre-selection of fingerlings and information on culled fish affect genetic evaluation of grow-out weight traits in two environments. Fish for different datasets were randomly sampled (R) or pre-selected and information on culled fish were either individually measured (S+IND), replicated with their family-specific averages (S+AVER), or missing (S-MIS). Variance estimates in R and S+IND did not diverge from simulated values, whereas S+AVER decreased residual variances. Accuracies of EBVs were equally high for R, S+IND and S+AVER. For S-MIS, convergence problems occurred, variance components were distorted, and EBV accuracies were low. Selection bias was consistently expressed by overestimated genetic and common environment variances, and underestimated residual variances. Data adjustment by S+AVER is concluded to sufficiently control for selection bias in genetic evaluation of growth, but for estimation of variances R or S+IND are preferable.

Keywords: fish breeding, selection bias, simulation