|
Abstract:
|
Complex diseases typically differ among individuals due to differences in genetic and environmental random effects. Studies with monozygotic and dizygotic twins allow us to fit the popular "ACE" model which estimates the proportion of trait variance explained by additive genetic (A), common environment (C), and unique environmental (E) latent effects, thus helping us better understand disease risk and etiology. The ACE model is typically fit to outcomes assumed to be normally distributed; however, lack of normality can lead to substantially erroneous results. In practice, we are often interested in estimating genetic and environmental random effects in highly non-normal traits, such as discrete counts or skewed continuous data. Therefore, we develop a flexible Generalized Method of Moments (GMM) framework for fitting ACE models in twin studies that requires no parametric distributional assumptions; rather only the first two moments need to be correctly specified. Through simulations and real data application, we show that the GMM ACE model provides a robust framework for estimating genetic and environmental variance components in both non-normal and approximately normal traits.
|
