MTDFREML

MTDFREML estimates genetic (co)variances in animal models using multiple-trait derivative-free restricted maximum likelihood (MTDFREML) to analyze correlated traits and complex model structures.

Key Features:

• Multiple-Trait Analysis: Handles multiple traits simultaneously to estimate correlated genetic effects and (co)variance components.
• Derivative-free REML: Uses a derivative-free restricted maximum likelihood approach that does not require computation of derivatives.
• Flexible Model Support: Supports arbitrary numbers of covariates, fixed effects, and independent random effects for each trait.
• Handling Missing Data: Allows any combination of missing traits within the multiple-trait framework.
• Empirical Consistency with MTGSAM: Demonstrated empirical unbiasedness and correlations between MTGSAM (with flat priors) and MTDFREML exceeding 0.99 in comparative studies.

Scientific Applications:

• Animal Breeding and Genetics: Estimation of genetic (co)variances for selection and breeding value evaluation in animal models.
• Variance Component Inference: Inference of genetic and residual covariance components in multiple-trait analyses.
• Analysis with Incomplete Phenotypes: Analysis of datasets with missing trait records and complex fixed/random effect structures.

Methodology:

Employs multiple-trait derivative-free restricted maximum likelihood to estimate variance components without computing derivatives; supports arbitrary covariates, fixed effects, independent random effects per trait, and any combination of missing traits; compared empirically with MTGSAM (Multiple-Trait Gibbs Sampling Algorithm) implemented in FORTRAN, which uses Gibbs sampling and informative priors, with reported correlations >0.99 versus MTDFREML using flat priors.