Simple Penalties on Maximum-Likelihood Estimates of Genetic Parameters to Reduce Sampling Variation

Area of Science:

• Quantitative genetics
• Statistical genetics
• Bioinformatics

Background:

• Multivariate genetic analyses are crucial for understanding complex traits.
• Estimating genetic covariances can be challenging due to substantial sampling variation, particularly with small datasets or numerous traits.
• Standard maximum-likelihood methods may yield imprecise estimates in such scenarios.

Purpose of the Study:

• To present a modified maximum-likelihood procedure for estimating genetic covariances.
• To enhance the precision of multivariate genetic parameter estimates by reducing sampling variances.
• To introduce a penalized likelihood approach with a default, data-derived penalty function.

Main Methods:

• A modification of standard maximum-likelihood procedures by incorporating a penalty term.
• Maximizing the likelihood function subject to a penalty derived from a Beta distribution of scale-free functions of covariance components.
• Extensive simulation studies to evaluate the performance of the penalized method.

Main Results:

• The penalized maximum-likelihood approach significantly reduces sampling variances in genetic covariance estimates.
• Mild, default penalties yield substantial reductions in loss (difference from population values) across various scenarios.
• The method effectively reduces loss without distorting phenotypic covariance estimates and performs well even in challenging cases.

Conclusions:

• Penalized maximum-likelihood provides a robust and efficient method for improving multivariate genetic parameter estimation.
• The proposed default penalty offers a practical solution to the challenge of determining penalization stringency.
• This statistical advancement has broad implications for quantitative genetics and related fields requiring precise genetic parameter estimates.