A1082
Title: Estimating signal-to-noise ratios for multivariate high-dimensional linear models
Authors: Xiaodong Li - UC Davis (United States) [presenting]
Abstract: Signal-to-noise ratios (SNR) are essential components in a wide range of statistical models, playing crucial roles in applications such as selecting tuning parameters for predictive models and estimating heritability in genomics. The method-of-moments estimator is a widely used approach for estimating SNR, primarily explored in single-response settings. The purpose is to extend the method-of-moments SNR estimation framework to linear models with multivariate responses. The approach encompasses both fixed-effects and random-effects models, and for each case, the asymptotic distributions of the proposed estimators are derived. Furthermore, scenarios involving non-Gaussian or heterogeneous noise are addressed by developing robust asymptotic inference procedures based on the theoretical findings. The effectiveness of the methods is demonstrated through extensive numerical experiments.
