Can a model with Genetic Groups for Mendelian Sampling Deviations Correct for pre-selection bias?

Pre-selection of animals in breeding schemes invalidates some of the assumptions necessary to maintain the unbiasedness property in genetic evaluations. In particular, the expected value of Mendelian sampling deviations is no longer zero, and the Mendelian sampling variance is reduced. This study investigated, using simulation, the possibility to attenuate bias due to preselection using a genetic evaluation model with genetic groups for Mendelian sampling deviations proposed some years ago. Two generations of data were simulated. Daughter yield deviations and yield deviations were simulated for males and females, resp., in generations 1 and 2. Males in generation 1 were pre-selected on estimated Mendelian sampling deviations. Accounting for preselection by including genetic groups for Mendelian sampling deviations did not remove bias in estimated breeding values. The reason was that pre-selection not only led to selection differential for Mendelian sampling deviations but also for parent averages, which the model failed to account for.