Supplemental Document for “Matrix Completion with Covariate Information”
Xiaojun Mao,Songxi Chen,R. K. Wong
2017-01-01
Abstract:This document provides supplementary material to the article “Matrix Completion with Covariate Information” written by the same authors. S1 Proof of Propositions Proof of Proposition 1. We have E ∥Xβ +B −W ◦Θ∗ ◦ Y ∥2F = ∥Xβ +B∥2F − 2 ⟨Xβ +B,E (W ◦Θ ∗ ◦ Y )⟩+ E ∥W ◦Θ∗ ◦ Y ∥2F = ∥(Xβ +B)− (Xβ0 +B0)∥2F − ∥Xβ0 +B0∥ 2 F + ∑ ij ( (Xβ0)ij +B0ij )2 + σ2 ij θij , due to Conditions C1(a) and C4. For any minimizer (βs,Bs) of R, we have Xβ0+B0 = Xβs+Bs, which implies X(β0−βs) = Bs−B0. Since Bs−B0 ∈ N (X), we can conclude both Xβs = Xβ0 ∗Xiaojun Mao is Ph.D. candidate, Department of Statistics, Iowa State University, Ames, IA 50011, USA (Email: mxjki@iastate.edu). †Author of Correspondence. Song Xi Chen is Chair Professor, Department of Business Statistics and Econometrics, Guanghua School of Management and Center for Statistical Science, Peking University, Beijing 100651, China (Email: csx@gsm.pku.edu.cn). His research is partially supported by Chinas National Key Research Special Program Grants 2016YFC0207701 and 2015CB856000, and National Natural Science Foundation of China grants 11131002, 71532001 and 71371016. ‡Raymond K. W. Wong is Assistant Professor, Department of Statistics, Texas A&M University, College Station, TX 77843, USA (Email: raywong@stat.tamu.edu). His research is partially supported by the National Science Foundation under Grants DMS-1612985 and DMS-1711952 (subcontract).