Related Experiment Videos
Sequential transformation for multiple traits for estimation of (co)variance components with a derivative-free
1Roman L. Hruska U.S. Meat Animal Research Center, ARS, USDA, Clay Center, NE 68933-0166.
Journal of Animal Science
|April 1, 1993
Summary
This study introduces a data transformation method for estimating covariance matrices in multiple-trait records. This approach simplifies mixed-model equations, enhancing computational efficiency and reducing memory requirements.
Area of Science:
- Quantitative Genetics
- Statistical Modeling
Background:
- Estimating covariance matrices is crucial for analyzing multiple-trait records in sequential selection.
- Derivative-free and derivative methods are commonly used, but can be computationally intensive.
Purpose of the Study:
- To present a data transformation technique for simplifying covariance matrix estimation in mixed models.
- To improve the efficiency of restricted likelihood maximization for multiple-trait records.
Main Methods:
- Data transformation using Choleski decomposition of the residual covariance matrix.
- Modification of mixed-model equations for iterative likelihood calculation.
- Utilizing derivative-free algorithms for restricted likelihood maximization.
Main Results:
- Transformed equations simplify mixed-model setup and iterative updates.
- The least squares component remains constant across rounds.
- A slight improvement in solution time (1-5%) was observed in an example.
Conclusions:
- Data transformation offers computational advantages for estimating covariance matrices in mixed models.
- The method simplifies equation management and reduces memory needs.
- This technique is beneficial for sequential selection analyses with multiple traits.