Multiple-trait restricted maximum likelihood for simulated measures of ovulation rate with underlying multivariate...

Area of Science:

• Quantitative Genetics
• Animal Breeding
• Statistical Genetics

Background:

• Estimating covariance components is crucial for animal breeding and genetic studies.
• Restricted Maximum Likelihood (REML) is a standard method for estimating these components.
• Ovulation rate, a key reproductive trait, is often analyzed using multiple measures across estrous cycles.

Purpose of the Study:

• To simulate multivariate normal traits and truncate them to binomial traits to mimic real biological data.
• To evaluate the accuracy of heritability and genetic correlation estimates derived from binomial traits compared to underlying normal traits.
• To assess the performance of multiple-trait REML in estimating genetic parameters for simulated ovulation rate data.

Main Methods:

• A template dataset with eight ovulation rate measures was used to simulate eight multivariate normal traits.
• Simulated normal traits were truncated to create binomial traits.
• Multiple-trait REML was applied to estimate heritabilities and genetic correlations from the simulated binomial data across various heritability and genetic correlation levels.

Main Results:

• Heritability estimates derived from the binomial scale consistently overestimated heritability on the normal scale.
• Genetic correlations estimated on the binomial scale significantly underestimated the true genetic correlations on the normal scale.
• Standard errors from replicated simulations were slightly larger than those predicted by the REMLPK software.

Conclusions:

• Standard transformations for heritability from binomial to normal scales can lead to overestimation.
• Estimating genetic correlations from binomial traits requires careful consideration due to substantial underestimation.
• The simulation results provide valuable insights into the biases and precision of genetic parameter estimation in animal models with binary traits.