|
Abstract:
|
Genome-wide association studies (GWAS) have identified thousands of genetic variants associated with complex traits. Among these discoveries, many complex traits are found to have shared genetic etiology. Genetic covariance is defined as the underlying covariance of genetic values and can be used to measure the shared genetic architecture. This paper proposes a unified method for robust estimation and inference procedure for genetic covariance of general outcomes that may be associated with high-dimensional genetic variants in nonlinear forms. The data of two outcomes may come from the same group or different groups of individuals. The proposed estimators are robust under a certain level of model mis-specification. %Another parameter of interest is the ``narrow-sense generic covariance'', which is a bilinear functional defined within linear working models. Our method based on linear working models provides robust inference for the narrow-sense generic covariance, even when both linear models are mis-specified. The performance of the method is evaluated in various numerical experiments to support the theoretical results.
|
