On the covariance between parameter estimates in models of twin data

Area of Science:

• Quantitative Genetics
• Behavioral Genetics
• Statistical Genetics

Background:

• Likelihood-based twin analyses are standard for dissecting genetic and environmental influences on traits.
• Understanding the covariance between variance component estimates is crucial for study design and interpretation.
• Previous work has not fully elucidated the interrelationships between additive genetic, dominance genetic, and common environmental variance estimates in twin studies.

Purpose of the Study:

• To investigate the covariance between estimates of additive genetic variance and dominance genetic variance or common environmental variance in twin analyses.
• To determine optimal sample size ratios for monozygotic (MZ) and dizygotic (DZ) twins.
• To assess the conditions under which variance components are efficiently estimable from twin data.

Main Methods:

• Utilized asymptotic covariances of variance component estimates for standard twin models.
• Calculated asymptotic correlations between additive genetic variance, dominance genetic variance, and common environmental variance estimates.
• Employed simulation studies to confirm analytical results and explore parameter estimability.

Main Results:

• Presented asymptotic covariance terms for commonly used twin models.
• Demonstrated how covariance terms inform optimal MZ:DZ sample size ratios.
• Highlighted that twin-only data inherently limits the ability to distinguish individual variance components, particularly additive genetic, dominance genetic, and common environmental factors.

Conclusions:

• Twin data alone imposes inherent limitations on the precise estimation of specific variance components.
• Additional family data beyond twins significantly improves the efficiency and power of genetic analyses.
• Understanding these statistical limitations is essential for accurate inference in genetic research using twin designs.