Needles and straw in a haystack: Robust confidence for possibly sparse sequences

Eduard Belitser, Nurzhan Nurushev

Research output: Contribution to JournalArticleAcademicpeer-review

Abstract

In the general signal+noise (allowing non-normal, non-independent observations) model, we construct an empirical Bayes posterior which we then use for uncertainty quantification for the unknown, possibly sparse, signal. We introduce a novel excessive bias restriction (EBR) condition, which gives rise to a new slicing of the entire space that is suitable for uncertainty quantification. Under EBR and some mild exchangeable exponential moment condition on the noise, we establish the local (oracle) optimality of the proposed confidence ball. Without EBR, we propose another confidence ball of full coverage, but its radius contains an additional σn1/4-term. In passing, we also get the local optimal results for estimation, posterior contraction problems, and the problem of weak recovery of sparsity structure. Adaptive minimax results (also for the estimation and posterior contraction problems) over various sparsity classes follow from our local results.

Original languageEnglish
Pages (from-to)191-225
Number of pages35
JournalBernoulli
Volume26
Issue number1
Early online date26 Nov 2019
DOIs
Publication statusPublished - 1 Jan 2020

Fingerprint

Confidence
Uncertainty Quantification
Restriction
Sparsity
Contraction
Ball
Empirical Bayes
Moment Conditions
Slicing
Minimax
Optimality
Coverage
Recovery
Radius
Entire
Unknown
Term
Model
Observation
Class

Keywords

  • Confidence set
  • Empirical Bayes posterior
  • Local radial rate

Cite this

@article{9fd5382bcc674b1ab197b4f720abc8ed,
title = "Needles and straw in a haystack: Robust confidence for possibly sparse sequences",
abstract = "In the general signal+noise (allowing non-normal, non-independent observations) model, we construct an empirical Bayes posterior which we then use for uncertainty quantification for the unknown, possibly sparse, signal. We introduce a novel excessive bias restriction (EBR) condition, which gives rise to a new slicing of the entire space that is suitable for uncertainty quantification. Under EBR and some mild exchangeable exponential moment condition on the noise, we establish the local (oracle) optimality of the proposed confidence ball. Without EBR, we propose another confidence ball of full coverage, but its radius contains an additional σn1/4-term. In passing, we also get the local optimal results for estimation, posterior contraction problems, and the problem of weak recovery of sparsity structure. Adaptive minimax results (also for the estimation and posterior contraction problems) over various sparsity classes follow from our local results.",
keywords = "Confidence set, Empirical Bayes posterior, Local radial rate",
author = "Eduard Belitser and Nurzhan Nurushev",
year = "2020",
month = "1",
day = "1",
doi = "10.3150/19-BEJ1122",
language = "English",
volume = "26",
pages = "191--225",
journal = "Bernoulli: A Journal of Mathematical Statistics and Probability",
issn = "1350-7265",
publisher = "International Statistical Institute",
number = "1",

}

Needles and straw in a haystack : Robust confidence for possibly sparse sequences. / Belitser, Eduard; Nurushev, Nurzhan.

In: Bernoulli, Vol. 26, No. 1, 01.01.2020, p. 191-225.

Research output: Contribution to JournalArticleAcademicpeer-review

TY - JOUR

T1 - Needles and straw in a haystack

T2 - Robust confidence for possibly sparse sequences

AU - Belitser, Eduard

AU - Nurushev, Nurzhan

PY - 2020/1/1

Y1 - 2020/1/1

N2 - In the general signal+noise (allowing non-normal, non-independent observations) model, we construct an empirical Bayes posterior which we then use for uncertainty quantification for the unknown, possibly sparse, signal. We introduce a novel excessive bias restriction (EBR) condition, which gives rise to a new slicing of the entire space that is suitable for uncertainty quantification. Under EBR and some mild exchangeable exponential moment condition on the noise, we establish the local (oracle) optimality of the proposed confidence ball. Without EBR, we propose another confidence ball of full coverage, but its radius contains an additional σn1/4-term. In passing, we also get the local optimal results for estimation, posterior contraction problems, and the problem of weak recovery of sparsity structure. Adaptive minimax results (also for the estimation and posterior contraction problems) over various sparsity classes follow from our local results.

AB - In the general signal+noise (allowing non-normal, non-independent observations) model, we construct an empirical Bayes posterior which we then use for uncertainty quantification for the unknown, possibly sparse, signal. We introduce a novel excessive bias restriction (EBR) condition, which gives rise to a new slicing of the entire space that is suitable for uncertainty quantification. Under EBR and some mild exchangeable exponential moment condition on the noise, we establish the local (oracle) optimality of the proposed confidence ball. Without EBR, we propose another confidence ball of full coverage, but its radius contains an additional σn1/4-term. In passing, we also get the local optimal results for estimation, posterior contraction problems, and the problem of weak recovery of sparsity structure. Adaptive minimax results (also for the estimation and posterior contraction problems) over various sparsity classes follow from our local results.

KW - Confidence set

KW - Empirical Bayes posterior

KW - Local radial rate

UR - http://www.scopus.com/inward/record.url?scp=85076545255&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85076545255&partnerID=8YFLogxK

U2 - 10.3150/19-BEJ1122

DO - 10.3150/19-BEJ1122

M3 - Article

VL - 26

SP - 191

EP - 225

JO - Bernoulli: A Journal of Mathematical Statistics and Probability

JF - Bernoulli: A Journal of Mathematical Statistics and Probability

SN - 1350-7265

IS - 1

ER -