Statistical properties of simple random-effects models for genetic heritability

Area of Science:

• Genetics
• Statistical Genetics
• Bioinformatics

Background:

• Random-effects models are widely used for estimating narrow-sense heritability (h²).
• Recent debates question the validity of conclusions drawn from these models using SNP data.
• Understanding the statistical properties of heritability estimates is crucial.

Purpose of the Study:

• To derive fundamental statistical properties of heritability estimates from random-effects models.
• To investigate model behavior when key assumptions are not met (e.g., shared environment, measurement error).
• To provide a baseline for variance and bias calculations in genetic analyses.

Main Methods:

• Derivation of statistical properties of heritability estimates.
• Manipulation of the score function for interpretable results.
• Exploration of model behavior under violated assumptions using genotype matrix singular values.

Main Results:

• Bias in heritability estimates from random-effects models is generally small.
• The score function can be adapted for clearer interpretation of results.
• Model variance and bias are critically dependent on genotype matrix singular value distributions.

Conclusions:

• Random-effects models provide a valid framework for heritability estimation with manageable bias.
• The study offers a method to interpret results and assess model robustness.
• A baseline for variance and bias is established for independent genotypes, aiding future genetic studies.