A0846
Title: Convergence rate of the James-Stein principal component
Authors: Youhong Lee - University of California, Santa Barbara (United States) [presenting]
Alexander Shkolnik - University of California, Santa Barbara (United States)
Abstract: The James-Stein estimator is presented for eigenvectors, building upon previous research that highlights its ability to reduce the distance to the true signal through the shrinkage constant and corrected eigenvector. Moving beyond the point estimates, the investigation focuses on the variability inherent to the James-Stein estimator. The sources of this variability are explored, and its impact is assessed on the estimator's components. The aim is to elucidate the asymptotic variance of the James-Stein estimator, particularly in relation to the signal-to-noise ratio and the shrinkage target. Moreover, the effect of James-Stein-type shrinkage is examined on this variability, differentiating between signal-noise and noise-noise correlations.
