Online Program Home
  My Program

Abstract Details

Activity Number: 73 - Nonparametric Statistics in High-Dimensional Settings
Type: Contributed
Date/Time: Sunday, July 30, 2017 : 4:00 PM to 5:50 PM
Sponsor: Section on Nonparametric Statistics
Abstract #323233 View Presentation
Title: Sparse Covariance Estimation via Concentration Inequalities
Author(s): Adam Kashlak* and Linglong Kong
Companies: Univ of Cambridge and University of Alberta
Keywords: sparse estimation ; covariance matrix ; concentration inequality ; high dimensional data

We introduce a new approach to the estimation of sparse covariance matrices for high dimensional data making use of concentration inequalities. The inequalities are used to a construct nonasymptotic confidence set about the empirical estimate. We then apply a search procedure to find an optimally sparse member of that set. This approach takes inspiration and can, in some sense, be thought of as both a penalized estimator and threshold estimator. The method is shown to have excellent performance when compared to both lasso and hard and soft threshold estimators for the covariance matrix.

Authors who are presenting talks have a * after their name.

Back to the full JSM 2017 program

Copyright © American Statistical Association