Comparison of methods for handling censored records in beef fertility data: simulation study

Area of Science:

• Animal Science
• Genetics
• Statistical Modeling

Background:

• Censored records are common in animal breeding data, particularly for traits like days to calving.
• Accurate handling of censored data is crucial for reliable genetic evaluations and breeding value predictions.

Purpose of the Study:

• To compare the performance of two distinct methods for handling censored records in beef cattle days to calving data.
• To evaluate the impact of different censoring rates (12% and 20%) on the accuracy of genetic parameter estimation and breeding value prediction.

Main Methods:

• A simulation study generated data mimicking real beef cattle populations, including genetic relationships.
• Two methods, Direct Censoring Penalty (DCPEN) and Direct Censoring Simulation (DCSIM), were applied to censored records.
• Bayesian inference using Gibbs sampling was employed to estimate variance components and breeding values.

Main Results:

• The DCSIM method showed less bias in residual variance estimation compared to DCPEN, as indicated by HPD intervals.
• Model comparison criteria (Bayes Factor, DIC) favored the DCSIM method.
• Despite statistical differences, both methods resulted in minimal reranking of sires, indicating similar practical outcomes for genetic evaluations.

Conclusions:

• The DCSIM method is statistically superior for handling censored calving interval data due to reduced bias.
• Both DCPEN and DCSIM methods provide comparable results for sire ranking in genetic evaluations for days to calving.
• The choice between methods may depend on specific data characteristics and desired precision in genetic parameter estimation.