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

Mathematics > Statistics Theory

Title: Tail-adaptive Bayesian shrinkage

Abstract: Modern genomic studies are increasingly focused on discovering more and more interesting genes associated with a health response. Traditional shrinkage priors are primarily designed to detect a handful of signals from tens and thousands of predictors. Under diverse sparsity regimes, the nature of signal detection is associated with a tail behaviour of a prior. A desirable tail behaviour is called tail-adaptive shrinkage property where tail-heaviness of a prior gets adaptively larger (or smaller) as a sparsity level increases (or decreases) to accommodate more (or less) signals. We propose a global-local-tail (GLT) Gaussian mixture distribution to ensure this property and provide accurate inference under diverse sparsity regimes. Incorporating a peaks-over-threshold method in extreme value theory, we develop an automated tail learning algorithm for the GLT prior. We compare the performance of the GLT prior to the Horseshoe in two gene expression datasets and numerical examples. Results suggest that varying tail rule is advantageous over fixed tail rule under diverse sparsity domains.
Subjects: Statistics Theory (math.ST); Applications (stat.AP); Computation (stat.CO); Methodology (stat.ME); Machine Learning (stat.ML)
Cite as: arXiv:2007.02192 [math.ST]
  (or arXiv:2007.02192v2 [math.ST] for this version)

Submission history

From: Se Yoon Lee [view email]
[v1] Sat, 4 Jul 2020 21:40:12 GMT (3396kb,D)
[v2] Wed, 13 Jan 2021 03:47:49 GMT (4103kb,D)

Link back to: arXiv, form interface, contact.