False Discovery Rate Smoothing

Wesley Tansey, Oluwasanmi Oluseye Koyejo, Russell A. Poldrack, James G. Scott

Research output: Contribution to journalArticle

Abstract

We present false discovery rate (FDR) smoothing, an empirical-Bayes method for exploiting spatial structure in large multiple-testing problems. FDR smoothing automatically finds spatially localized regions of significant test statistics. It then relaxes the threshold of statistical significance within these regions, and tightens it elsewhere, in a manner that controls the overall false discovery rate at a given level. This results in increased power and cleaner spatial separation of signals from noise. The approach requires solving a nonstandard high-dimensional optimization problem, for which an efficient augmented-Lagrangian algorithm is presented. In simulation studies, FDR smoothing exhibits state-of-the-art performance at modest computational cost. In particular, it is shown to be far more robust than existing methods for spatially dependent multiple testing. We also apply the method to a dataset from an fMRI experiment on spatial working memory, where it detects patterns that are much more biologically plausible than those detected by standard FDR-controlling methods. All code for FDR smoothing is publicly available in Python and R (https://github.com/tansey/smoothfdr). Supplementary materials for this article are available online.

Original languageEnglish (US)
Pages (from-to)1156-1171
Number of pages16
JournalJournal of the American Statistical Association
Volume113
Issue number523
DOIs
StatePublished - Jul 3 2018

Fingerprint

Smoothing
Multiple Testing
Augmented Lagrangians
Empirical Bayes Method
Working Memory
Python
Augmented Lagrangian
Functional Magnetic Resonance Imaging
Statistical Significance
Spatial Structure
Test Statistic
False
Computational Cost
High-dimensional
Simulation Study
Optimization Problem
Dependent
Experiment
Testing

Keywords

  • Empirical Bayes
  • FDR
  • False discovery rate
  • Fused lasso
  • Multiple hypothesis testing
  • Spatial smoothing

ASJC Scopus subject areas

  • Statistics and Probability
  • Statistics, Probability and Uncertainty

Cite this

False Discovery Rate Smoothing. / Tansey, Wesley; Koyejo, Oluwasanmi Oluseye; Poldrack, Russell A.; Scott, James G.

In: Journal of the American Statistical Association, Vol. 113, No. 523, 03.07.2018, p. 1156-1171.

Research output: Contribution to journalArticle

Tansey, Wesley ; Koyejo, Oluwasanmi Oluseye ; Poldrack, Russell A. ; Scott, James G. / False Discovery Rate Smoothing. In: Journal of the American Statistical Association. 2018 ; Vol. 113, No. 523. pp. 1156-1171.
@article{85f64d685af7483abd13ba6fa3a0a305,
title = "False Discovery Rate Smoothing",
abstract = "We present false discovery rate (FDR) smoothing, an empirical-Bayes method for exploiting spatial structure in large multiple-testing problems. FDR smoothing automatically finds spatially localized regions of significant test statistics. It then relaxes the threshold of statistical significance within these regions, and tightens it elsewhere, in a manner that controls the overall false discovery rate at a given level. This results in increased power and cleaner spatial separation of signals from noise. The approach requires solving a nonstandard high-dimensional optimization problem, for which an efficient augmented-Lagrangian algorithm is presented. In simulation studies, FDR smoothing exhibits state-of-the-art performance at modest computational cost. In particular, it is shown to be far more robust than existing methods for spatially dependent multiple testing. We also apply the method to a dataset from an fMRI experiment on spatial working memory, where it detects patterns that are much more biologically plausible than those detected by standard FDR-controlling methods. All code for FDR smoothing is publicly available in Python and R (https://github.com/tansey/smoothfdr). Supplementary materials for this article are available online.",
keywords = "Empirical Bayes, FDR, False discovery rate, Fused lasso, Multiple hypothesis testing, Spatial smoothing",
author = "Wesley Tansey and Koyejo, {Oluwasanmi Oluseye} and Poldrack, {Russell A.} and Scott, {James G.}",
year = "2018",
month = "7",
day = "3",
doi = "10.1080/01621459.2017.1319838",
language = "English (US)",
volume = "113",
pages = "1156--1171",
journal = "Journal of the American Statistical Association",
issn = "0162-1459",
publisher = "Taylor & Francis",
number = "523",

}

TY - JOUR

T1 - False Discovery Rate Smoothing

AU - Tansey, Wesley

AU - Koyejo, Oluwasanmi Oluseye

AU - Poldrack, Russell A.

AU - Scott, James G.

PY - 2018/7/3

Y1 - 2018/7/3

N2 - We present false discovery rate (FDR) smoothing, an empirical-Bayes method for exploiting spatial structure in large multiple-testing problems. FDR smoothing automatically finds spatially localized regions of significant test statistics. It then relaxes the threshold of statistical significance within these regions, and tightens it elsewhere, in a manner that controls the overall false discovery rate at a given level. This results in increased power and cleaner spatial separation of signals from noise. The approach requires solving a nonstandard high-dimensional optimization problem, for which an efficient augmented-Lagrangian algorithm is presented. In simulation studies, FDR smoothing exhibits state-of-the-art performance at modest computational cost. In particular, it is shown to be far more robust than existing methods for spatially dependent multiple testing. We also apply the method to a dataset from an fMRI experiment on spatial working memory, where it detects patterns that are much more biologically plausible than those detected by standard FDR-controlling methods. All code for FDR smoothing is publicly available in Python and R (https://github.com/tansey/smoothfdr). Supplementary materials for this article are available online.

AB - We present false discovery rate (FDR) smoothing, an empirical-Bayes method for exploiting spatial structure in large multiple-testing problems. FDR smoothing automatically finds spatially localized regions of significant test statistics. It then relaxes the threshold of statistical significance within these regions, and tightens it elsewhere, in a manner that controls the overall false discovery rate at a given level. This results in increased power and cleaner spatial separation of signals from noise. The approach requires solving a nonstandard high-dimensional optimization problem, for which an efficient augmented-Lagrangian algorithm is presented. In simulation studies, FDR smoothing exhibits state-of-the-art performance at modest computational cost. In particular, it is shown to be far more robust than existing methods for spatially dependent multiple testing. We also apply the method to a dataset from an fMRI experiment on spatial working memory, where it detects patterns that are much more biologically plausible than those detected by standard FDR-controlling methods. All code for FDR smoothing is publicly available in Python and R (https://github.com/tansey/smoothfdr). Supplementary materials for this article are available online.

KW - Empirical Bayes

KW - FDR

KW - False discovery rate

KW - Fused lasso

KW - Multiple hypothesis testing

KW - Spatial smoothing

UR - http://www.scopus.com/inward/record.url?scp=85048057764&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85048057764&partnerID=8YFLogxK

U2 - 10.1080/01621459.2017.1319838

DO - 10.1080/01621459.2017.1319838

M3 - Article

VL - 113

SP - 1156

EP - 1171

JO - Journal of the American Statistical Association

JF - Journal of the American Statistical Association

SN - 0162-1459

IS - 523

ER -