A0986
Title: Variance estimation for multivariate high-dimensional random effects models under heteroskedasticity
Authors: Xiaodong Li - UC Davis (United States) [presenting]
Xiaohan Hu - UC Davis (United States)
Zhentao Li - UC Davis (United States)
Abstract: Variance estimation in high-dimensional random effects models has recently been widely used in genomics for model-based heritability estimation, and extensions to multivariate traits have also attracted much attention in various phenotypically rich studies. The purpose is to introduce the recent work on making inferences about certain variance parameters, e.g. signal-to-noise ratios, in high-dimensional random effects models with multivariate responses under heteroskedasticity. Two methods are considered: a method of moments and a likelihood-based aggregated estimating equation method. For each method, the consistency and asymptotic distribution of the estimator is established. In particular, the results characterize how the standard errors of the estimators depend on the multivariate noise heteroskedasticity.
