Influence of partitioning data by sex on genetic variance and covariance components for weaning weight in beef cattle.

Heterogeneity of (co)variance components (VC) by sex is currently accounted for in national genetic evaluations for Simmental cattle. Parameters used in the national evaluation program are estimated from data split into male, female, and steer populations. Analyzing only male data does not account for selection of females, and vice versa. To determine the impact of selection, a Monte Carlo simulation was used, and estimates of VC for weaning weight were obtained when data were partitioned by sex. Weaning weight data were simulated using homogeneous VC for males and females for random and selected populations. Restricted maximum likelihood estimates were obtained for direct and maternal genetic and permanent and temporary environmental variances and genetic covariance between direct and maternal effects by analyzing complete or split data. Estimates differed (P < .01) from input values in data from selected populations split by sex, yielding a spurious heterogeneity of VC for sex. The heterogeneity was reduced in models using genetic groups but not completely removed. Splitting data by sex also influenced VC estimates in data simulated with heterogeneous VC for males and females.