Simulation study on Mendelian sampling variance tests

Abstract

To be able to assess the quality of the data sets and national evaluation models, Interbull and national evaluation centers need a validation test. Thus far, two Mendelian sampling variance test methods have been proposed: Mendelian sampling (IB4) and full model sampling (FMS) variance estimation methods, but neither has been implemented. The aim of this simulation study was to dissect the behaviour of both methods under two different scenarios for bulls and cows. Scenario A served as a control that should pass the test. For Scenario B, a yearly increase of 2% in phenotypic variance was generated. Without heterogeneous variance adjustment, it should fail the test. As an alternative, an analysis of MACE model residuals could be a simple tool to check the data quality. On average, a yearly increase of 1.9% and 1.4% in genetic variance were observed for cows and bulls in Scenario B without HV adjustment. The IB4 test performed well when applied to cows and it was able to detect the simulated heterogeneity in genetic variance. A yearly increase of 1.3% in the variance of MACE residuals was observed in Scenario B without HV adjustment. This was consistent with the genetic variance estimates for bulls, indicating that the analysis of MACE residuals could be utilized to check the data quality for bulls.