Related Experiment Videos
A bootstrap approach to confidence regions for genetic parameters from Method R estimates
A Reverter1, C J Kaiser, C H Mallinckrodt
1Animal Genetics and Breeding Unit, University of New England, Armidale, NSW, Australia.
Journal of Animal Science
|October 22, 1998
Summary
Bootstrapping techniques provide reliable confidence regions (CR) for heritability (h2) and permanent environmental effects (c2) even with limited Method R estimates. This method ensures accurate genetic parameter estimation in animal models.
Area of Science:
- Quantitative Genetics
- Animal Breeding
- Statistical Genetics
Background:
- Estimating heritability (h2) and permanent environmental effects (c2) is crucial in animal breeding.
- Method R provides estimates, but computational demands can limit the number of replications.
- Assessing the accuracy of confidence regions (CR) for these estimates is essential.
Purpose of the Study:
- To evaluate bootstrapping techniques for generating confidence regions (CR) for heritability (h2) and permanent environmental effects (c2) using Method R estimates.
- To determine the optimal number of subsamples (NUMEST) required for accurate CR estimation.
- To compare the accuracy of parametric and nonparametric approaches for CR assessment.
Main Methods:
- Simulated data from a univariate, repeated measures, full animal model with 50% subsampling.
- Application of bootstrapping techniques to Method R estimates of h2 and c2.
- Assessment of CR accuracy using parametric (bivariate normality) and nonparametric (rank deviations) methods.
- Evaluation of average loss of confidence (LOSS) based on the number of estimates sampled (NUMEST).
Main Results:
- Bootstrap estimates of h2 and c2 converged rapidly, within 10(-3) of asymptotic values at NUMEST = 5.
- Similar convergence for standard error (SE) estimates was achieved with NUMEST = 20.
- Nonparametric CR were more accurate than parametric CR at NUMEST < 10.
- Parametric CR showed a faster rate of loss reduction with increasing NUMEST.
Conclusions:
- Bootstrapping techniques enable reliable confidence region (CR) estimation for heritability (h2) and permanent environmental effects (c2) even with a limited number of Method R estimates (e.g., 10-20).
- This approach offers a computationally feasible solution for obtaining accurate genetic parameter estimates in complex animal models.
- Both parametric and nonparametric methods can be employed, with the choice potentially depending on the number of available estimates.