Treating small herds as fixed or random in an animal model.

Area of Science:

• Animal breeding and genetics
• Quantitative genetics
• Statistical modeling in animal populations

Background:

• Animal breeding programs often involve small herd sizes and non-random associations between sires and herds.
• Preferential treatment of animals, influenced by sire genetic merit, can introduce bias in genetic evaluations.
• The choice of modeling herd effects (fixed vs. random) impacts the robustness of prediction models.

Purpose of the Study:

• To compare the predictive performance of animal models with fixed versus random herd effects.
• To evaluate the robustness of these models under conditions of small herd sizes and sire-herd non-randomness.
• To determine the impact of sire's genetic advantage on the accuracy of prediction models.

Main Methods:

• Simulated a population with small herd sizes and preferential treatment, modeled as non-random sire-herd association or sire's genetic advantage.
• Compared prediction robustness between animal models incorporating fixed herd effects and random herd effects.
• Assessed model accuracy and empirical mean squared error (MSE) under varying conditions, including selection.

Main Results:

• No difference in prediction accuracy between fixed and random herd models was observed for large herds.
• For small herds, the random herd model demonstrated higher accuracy and lower empirical MSE than the fixed herd model.
• The superiority of the random herd model in small herds persisted under selection, but its MSE increased with large sire effects, indicating potential bias.

Conclusions:

• The random herd model is generally superior to the fixed herd model for genetic evaluations involving small herds.
• While large sire effects can introduce bias in random herd models, such effects are considered unrealistic in practice.
• The findings support the use of random herd effects in animal models for more accurate genetic predictions in typical scenarios.