Model determination in a case of heterogeneity of variance using sampling techniques

Area of Science:

• Animal breeding and genetics
• Statistical modeling
• Quantitative genetics

Background:

• Heterogeneity of variance is a common issue in animal breeding data.
• Accurate statistical models are crucial for estimating genetic parameters like heritability.
• Previous methods may not adequately address sex-specific variance differences.

Purpose of the Study:

• To describe a sampling determination procedure for model comparison in cases of heterogeneous variance.
• To compare the performance of Single Trait, Heterogeneous Variance, and Multiple Trait Animal Models.
• To estimate heritability for weight at 210 days in Spanish Pirenaica cattle.

Main Methods:

• Utilized predictive distributions and Gibbs Sampling for computation.
• Employed Mean Square Error between predictive distribution expectations and real data as the comparison criterion.
• Applied the procedure to a dataset of 210-day weight in Spanish Pirenaica beef cattle.

Main Results:

• The Heterogeneous Variance Animal Model was identified as the most adjusted model.
• This suggests a balance between model complexity and data availability.
• Estimated heritabilities were 0.489 ± 0.076 for males and 0.331 ± 0.082 for females.

Conclusions:

• The described sampling determination procedure is effective for selecting appropriate models with heterogeneous variances.
• The Heterogeneous Variance Animal Model provides a more accurate representation for this specific dataset.
• Accurate heritability estimates are essential for effective genetic improvement programs in cattle breeding.