Session: Keynote Address I

Date & Time: Friday, May 22 — 9:45 AM–10:45 AM

Talk Title

Estimating covariance in high-dimensional random effect models

Abstract

With independent observations that are highly multivariate, a large literature studies estimation of covariance, addressing phenomena such as eigenvalue spreading and bias. As the samples are independent, one can speak of a single level of variation.

Random effects models, for example as used in quantitative genetics, introduce multiple levels of variation. When the measured phenotypes are highly multivariate, interest again turns to estimation of the covariance of the effects, and in particular aspects of the genetic covariance matrix.

The talk describes recent work in the random effect setting from both parametric and nonparametric perspectives. Key ingredients are the Marchenko-Pastur system of equations and a spectral central limit theorem developed for this setting. This is joint work with Ran Xie and Yuchen Wu.