The beta-binomial model for estimating heritabilities of binary traits.

Area of Science:

• Quantitative Genetics
• Statistical Genetics
• Animal Breeding

Background:

• Estimating heritability for binary traits is crucial in genetic studies.
• Existing methods, like the mixed linear threshold model, have limitations in accuracy.
• Stochastic simulation in randomized block experiments is a common approach.

Purpose of the Study:

• To propose and evaluate a novel beta-binomial model for estimating heritability.
• To compare the performance of the beta-binomial model against existing methods.
• To assess the model's utility in estimating selection response and intraclass correlations.

Main Methods:

• Utilized stochastic simulations of sib trials in completely randomized block designs.
• Applied a proposed beta-binomial model as a conjugate prior for trait probabilities.
• Linked the beta-binomial model to a threshold model for individual observations.

Main Results:

• The beta-binomial model demonstrated improved heritability estimates compared to the mixed linear threshold model.
• Average bias in beta-binomial estimates for heritability and family mean heritability was below 2%.
• Significant improvements were observed for heritabilities above 0.3 and low trait probabilities (p < 0.30).

Conclusions:

• The beta-binomial model offers a more accurate framework for heritability estimation in binary traits.
• This model facilitates the estimation of selection response and reliable family mean heritability.
• The beta-binomial approach allows direct estimation of intraclass correlations.