|
Activity Number:
|
166
- New Developments in High-Dimensional Statistics
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, July 31, 2017 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
IMS
|
|
Abstract #323250
|
View Presentation
|
|
Title:
|
Adaptive Sparse Estimation with Side Information
|
|
Author(s):
|
Trambak Banerjee* and Gourab Mukherjee and Wenguang Sun
|
|
Companies:
|
USC and University of Southern California and University of Southern California
|
|
Keywords:
|
Minimax ;
Sparsity ;
Adaptive ;
Covariate Assisted ;
Side Information ;
SURE
|
|
Abstract:
|
We consider the problem of estimating a sparse Gaussian mean vector when we also have side information from a covariate. We propose an efficient, adaptive methodology for leveraging the extra information in the covariate. Our proposed estimator not only adapts to the unknown sparsity of the signal but also to the degree of informativeness of our covariate. Over a wide range of scenarios, we exactly characterize the additional benefits of having side information given sparsity prior information on the parametric space and exhibit regimes where our proposed estimator far outperforms state-of-the art non-linear shrinkage estimators that do not use any covariate information. We assess optimal ways of using the covariate information and establish asymptotic minimaxity of our proposed estimator. The benefits of using side information via our method is illustrated on a concrete study in single cell virology where infection data from related strains of the virus were used as side information for studying the effects of the concerned strain.
|
Authors who are presenting talks have a * after their name.
