B0469
Title: Dimension reduction for tensor response regression models
Authors: Chung Eun Lee - Baruch College (United States) [presenting]
Xin Zhang - Florida State University (United States)
Lexin Li - University of California Berkeley (United States)
Abstract: A flexible model-free approach to the regression analysis of a tensor response and a vector predictor is proposed. Without specifying the specific form of the regression mean function, the estimation of the dimension reduction subspace that captures all the variations in the regression mean function is considered. A new nonparametric metric called tensor martingale difference divergence is proposed, and its statistical properties are studied. Built on this new metric, computationally efficient estimation and asymptotically valid procedures are developed. The method's efficacy through simulations and a real data application for e-commerce is demonstrated.
