We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:


Current browse context:


Change to browse by:

References & Citations


(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Statistics > Methodology

Title: A Compound Decision Approach to Covariance Matrix Estimation

Abstract: Covariance matrix estimation is a fundamental statistical task in many applications, but the sample covariance matrix is sub-optimal when the sample size is comparable to or less than the number of features. Such high-dimensional settings are common in modern genomics, where covariance matrix estimation is frequently employed as a method for inferring gene networks. To achieve estimation accuracy in these settings, existing methods typically either assume that the population covariance matrix has some particular structure, for example sparsity, or apply shrinkage to better estimate the population eigenvalues. In this paper, we study a new approach to estimating high-dimensional covariance matrices. We first frame covariance matrix estimation as a compound decision problem. This motivates defining a class of decision rules and using a nonparametric empirical Bayes g-modeling approach to estimate the optimal rule in the class. Simulation results and gene network inference in an RNA-seq experiment in mouse show that our approach is comparable to or can outperform a number of state-of-the-art proposals.
Comments: 20 pages, 4 figures. Biometrics (2022)
Subjects: Methodology (stat.ME); Statistics Theory (math.ST)
MSC classes: 62C12 (Primary) 62C25 (Secondary)
DOI: 10.1111/biom.13686
Cite as: arXiv:2005.04549 [stat.ME]
  (or arXiv:2005.04549v4 [stat.ME] for this version)

Submission history

From: Huiqin Xin [view email]
[v1] Sun, 10 May 2020 02:05:11 GMT (664kb,D)
[v2] Sat, 10 Apr 2021 23:00:58 GMT (294kb,D)
[v3] Fri, 24 Dec 2021 06:28:23 GMT (502kb,D)
[v4] Thu, 2 Jun 2022 23:08:49 GMT (273kb,D)

Link back to: arXiv, form interface, contact.