Simulation study on prior sensitivity in Bayesian variance component and heritability estimation for hierarchical breeding designs

Accurate estimation of variance components and heritability is fundamental in quantitative genetics and animal breeding. Bayesian methods based on linear mixed models offer a flexible framework for this purpose, but the choice of prior distribution for the variance components can substantially influence the resulting estimates — particularly when sample sizes are small or heritability is low. This study systematically investigates the sensitivity of Bayesian variance component and heritability estimates to the choice of prior distribution in a balanced nested sire-dam design. Several commonly used prior specifications, ranging from classical non-informative to modern weakly informative priors, are compared within a comprehensive Monte Carlo simulation framework. The simulation design varies key experimental factors, including the number of families, family sizes, and the level of true heritability, to cover a wide range of practically relevant scenarios. Estimation performance is evaluated using standard criteria such as bias, precision, interval coverage, and interval width. The findings aim to provide practical guidance for researchers on selecting appropriate prior distributions for variance component estimation in hierarchical breeding designs, with particular emphasis on conditions where prior influence is most pronounced.

Scientists involved

• Dr René Mauer
• Prof Dr Ingo Röder

Cooperation Partner

• Dr Ebru Kaya Başar (Akdeniz University, Antalya)