The benefit of group sparsity in group inference with de-biased scaled group Lasso

Ritwik Mitra, Cun Hui Zhang

Research output: Contribution to journalArticle

9 Citations (Scopus)

Abstract

We study confidence regions and approximate chi-squared tests for variable groups in high-dimensional linear regression. When the size of the group is small, low-dimensional projection estimators for individual coefficients can be directly used to construct efficient confidence regions and p-values for the group. However, the existing analyses of low-dimensional projection estimators do not directly carry through for chi-squared-based inference of a large group of variables without inflating the sample size by a factor of the group size. We propose to de-bias a scaled group Lasso for chi-squared-based statistical inference for potentially very large groups of variables. We prove that the proposed methods capture the benefit of group sparsity under proper conditions, for statistical inference of the noise level and variable groups, large and small. Such benefit is especially strong when the group size is large.

Original languageEnglish (US)
Pages (from-to)1829-1873
Number of pages45
JournalElectronic Journal of Statistics
Volume10
Issue number2
DOIs
StatePublished - Jan 1 2016

Fingerprint

Lasso
Sparsity
Biased
Projection Estimator
Chi-squared
Confidence Region
Statistical Inference
Inference
Chi-squared test
Large groups
p-Value
Linear regression
Sample Size
High-dimensional
Statistical inference
Group size
Confidence region
Estimator

All Science Journal Classification (ASJC) codes

  • Statistics and Probability
  • Statistics, Probability and Uncertainty

Keywords

  • Asymptotic normality
  • Bias correction
  • Chi-squared distribution
  • Group inference
  • Relaxed projection
  • Relaxed projection

Cite this

@article{5b82bd94805f4758b3d4b4a0cc3af7cc,
title = "The benefit of group sparsity in group inference with de-biased scaled group Lasso",
abstract = "We study confidence regions and approximate chi-squared tests for variable groups in high-dimensional linear regression. When the size of the group is small, low-dimensional projection estimators for individual coefficients can be directly used to construct efficient confidence regions and p-values for the group. However, the existing analyses of low-dimensional projection estimators do not directly carry through for chi-squared-based inference of a large group of variables without inflating the sample size by a factor of the group size. We propose to de-bias a scaled group Lasso for chi-squared-based statistical inference for potentially very large groups of variables. We prove that the proposed methods capture the benefit of group sparsity under proper conditions, for statistical inference of the noise level and variable groups, large and small. Such benefit is especially strong when the group size is large.",
keywords = "Asymptotic normality, Bias correction, Chi-squared distribution, Group inference, Relaxed projection, Relaxed projection",
author = "Ritwik Mitra and Zhang, {Cun Hui}",
year = "2016",
month = "1",
day = "1",
doi = "10.1214/16-EJS1120",
language = "English (US)",
volume = "10",
pages = "1829--1873",
journal = "Electronic Journal of Statistics",
issn = "1935-7524",
publisher = "Institute of Mathematical Statistics",
number = "2",

}

The benefit of group sparsity in group inference with de-biased scaled group Lasso. / Mitra, Ritwik; Zhang, Cun Hui.

In: Electronic Journal of Statistics, Vol. 10, No. 2, 01.01.2016, p. 1829-1873.

Research output: Contribution to journalArticle

TY - JOUR

T1 - The benefit of group sparsity in group inference with de-biased scaled group Lasso

AU - Mitra, Ritwik

AU - Zhang, Cun Hui

PY - 2016/1/1

Y1 - 2016/1/1

N2 - We study confidence regions and approximate chi-squared tests for variable groups in high-dimensional linear regression. When the size of the group is small, low-dimensional projection estimators for individual coefficients can be directly used to construct efficient confidence regions and p-values for the group. However, the existing analyses of low-dimensional projection estimators do not directly carry through for chi-squared-based inference of a large group of variables without inflating the sample size by a factor of the group size. We propose to de-bias a scaled group Lasso for chi-squared-based statistical inference for potentially very large groups of variables. We prove that the proposed methods capture the benefit of group sparsity under proper conditions, for statistical inference of the noise level and variable groups, large and small. Such benefit is especially strong when the group size is large.

AB - We study confidence regions and approximate chi-squared tests for variable groups in high-dimensional linear regression. When the size of the group is small, low-dimensional projection estimators for individual coefficients can be directly used to construct efficient confidence regions and p-values for the group. However, the existing analyses of low-dimensional projection estimators do not directly carry through for chi-squared-based inference of a large group of variables without inflating the sample size by a factor of the group size. We propose to de-bias a scaled group Lasso for chi-squared-based statistical inference for potentially very large groups of variables. We prove that the proposed methods capture the benefit of group sparsity under proper conditions, for statistical inference of the noise level and variable groups, large and small. Such benefit is especially strong when the group size is large.

KW - Asymptotic normality

KW - Bias correction

KW - Chi-squared distribution

KW - Group inference

KW - Relaxed projection

KW - Relaxed projection

UR - http://www.scopus.com/inward/record.url?scp=84978716180&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84978716180&partnerID=8YFLogxK

U2 - 10.1214/16-EJS1120

DO - 10.1214/16-EJS1120

M3 - Article

AN - SCOPUS:84978716180

VL - 10

SP - 1829

EP - 1873

JO - Electronic Journal of Statistics

JF - Electronic Journal of Statistics

SN - 1935-7524

IS - 2

ER -