A robust and unified framework for estimating heritability in twin studies using generalized estimating equations.

Area of Science:

• Quantitative genetics
• Biometrical genetics
• Statistical genetics

Background:

• Heritability quantifies genetic contributions to trait variance in populations.
• Twin studies are a primary method for estimating heritability, with over 17,000 traits studied.
• Traditional methods (NACE, Falconer's) struggle with non-normally distributed outcomes (binary, counts, skewed).

Purpose of the Study:

• To introduce a robust generalized estimating equations (GEE2) framework for heritability estimation.
• To extend heritability estimation to non-normally distributed traits.
• To develop an improved method, GEE2-Falconer, that can incorporate covariates.

Main Methods:

• Developed a unified GEE2 framework encompassing traditional methods.
• Proposed GEE2-Falconer for heritability estimation in non-normal data.
• Incorporated mean and variance-level covariate effects into the GEE2-Falconer model.

Main Results:

• GEE2 models demonstrate superior coverage of true heritability for non-normal outcomes compared to traditional approaches.
• The GEE2 framework provides robust standard errors.
• GEE2-Falconer allows heritability to vary with covariates like sex or age, unlike traditional Falconer's method.

Conclusions:

• GEE2-Falconer is recommended for accurate heritability estimation in twin studies with non-normally distributed outcomes.
• The GEE2 framework offers a unified and robust approach to heritability estimation.
• This method addresses limitations of traditional models, ensuring unbiased estimates even with non-normal data.