Sparse Hotelling's T^2 Test for Data from Two Samples

• Activity Number: 332
• Type: Contributed
• Date/Time: Tuesday, August 5, 2014 : 10:30 AM to 12:20 PM
• Sponsor: IMS
• Author(s): Subhajit Dutta*+ and Line H. Clemmensen and Marc G. Genton
• Companies: Spatio-Temporal Statistics and Data Analysis and Technical University of Denmark and King Abdullah University of Science and Technology
• Keywords: cross-validation ; elastic nets ; sparse data ; sub-sampling

Abstract

We study the problem of detecting a shift between the means of two multivariate normal distributions in the sparse setting. More specifically, we propose a test statistic that considers an adaptive projection, and then carries out the Hotelling's T^2 using the Moore-Penrose inverse of the covariance matrix. Power curves are generated from simulated data of varying size and difficulty, and compared with other state-of-the-art tests. We further support our claim using some preliminary theoretical results.