Testing a single regression coefficient in high dimensional linear models

Wei Lan, Ping Shou Zhong, Runze Li, Hansheng Wang, Chih Ling Tsai

Research output: Contribution to journalArticle

3 Citations (Scopus)

Abstract

In linear regression models with high dimensional data, the classical z-test (or t-test) for testing the significance of each single regression coefficient is no longer applicable. This is mainly because the number of covariates exceeds the sample size. In this paper, we propose a simple and novel alternative by introducing the Correlated Predictors Screening (CPS) method to control for predictors that are highly correlated with the target covariate. Accordingly, the classical ordinary least squares approach can be employed to estimate the regression coefficient associated with the target covariate. In addition, we demonstrate that the resulting estimator is consistent and asymptotically normal even if the random errors are heteroscedastic. This enables us to apply the z-test to assess the significance of each covariate. Based on the p-value obtained from testing the significance of each covariate, we further conduct multiple hypothesis testing by controlling the false discovery rate at the nominal level. Then, we show that the multiple hypothesis testing achieves consistent model selection. Simulation studies and empirical examples are presented to illustrate the finite sample performance and the usefulness of the proposed method, respectively.

Original languageEnglish (US)
Pages (from-to)154-168
Number of pages15
JournalJournal of Econometrics
Volume195
Issue number1
DOIs
StatePublished - Nov 1 2016

Fingerprint

Regression Coefficient
Covariates
Linear Model
High-dimensional
z-test
Testing
Multiple Hypothesis Testing
Predictors
Random errors
Linear regression
Target
Ordinary Least Squares
Random Error
t-test
High-dimensional Data
p-Value
Screening
Linear Regression Model
Model Selection
Categorical or nominal

All Science Journal Classification (ASJC) codes

  • Economics and Econometrics

Cite this

Lan, Wei ; Zhong, Ping Shou ; Li, Runze ; Wang, Hansheng ; Tsai, Chih Ling. / Testing a single regression coefficient in high dimensional linear models. In: Journal of Econometrics. 2016 ; Vol. 195, No. 1. pp. 154-168.
@article{9e06678d93204406aaca884d3d6e47f0,
title = "Testing a single regression coefficient in high dimensional linear models",
abstract = "In linear regression models with high dimensional data, the classical z-test (or t-test) for testing the significance of each single regression coefficient is no longer applicable. This is mainly because the number of covariates exceeds the sample size. In this paper, we propose a simple and novel alternative by introducing the Correlated Predictors Screening (CPS) method to control for predictors that are highly correlated with the target covariate. Accordingly, the classical ordinary least squares approach can be employed to estimate the regression coefficient associated with the target covariate. In addition, we demonstrate that the resulting estimator is consistent and asymptotically normal even if the random errors are heteroscedastic. This enables us to apply the z-test to assess the significance of each covariate. Based on the p-value obtained from testing the significance of each covariate, we further conduct multiple hypothesis testing by controlling the false discovery rate at the nominal level. Then, we show that the multiple hypothesis testing achieves consistent model selection. Simulation studies and empirical examples are presented to illustrate the finite sample performance and the usefulness of the proposed method, respectively.",
author = "Wei Lan and Zhong, {Ping Shou} and Runze Li and Hansheng Wang and Tsai, {Chih Ling}",
year = "2016",
month = "11",
day = "1",
doi = "10.1016/j.jeconom.2016.05.016",
language = "English (US)",
volume = "195",
pages = "154--168",
journal = "Journal of Econometrics",
issn = "0304-4076",
publisher = "Elsevier BV",
number = "1",

}

Testing a single regression coefficient in high dimensional linear models. / Lan, Wei; Zhong, Ping Shou; Li, Runze; Wang, Hansheng; Tsai, Chih Ling.

In: Journal of Econometrics, Vol. 195, No. 1, 01.11.2016, p. 154-168.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Testing a single regression coefficient in high dimensional linear models

AU - Lan, Wei

AU - Zhong, Ping Shou

AU - Li, Runze

AU - Wang, Hansheng

AU - Tsai, Chih Ling

PY - 2016/11/1

Y1 - 2016/11/1

N2 - In linear regression models with high dimensional data, the classical z-test (or t-test) for testing the significance of each single regression coefficient is no longer applicable. This is mainly because the number of covariates exceeds the sample size. In this paper, we propose a simple and novel alternative by introducing the Correlated Predictors Screening (CPS) method to control for predictors that are highly correlated with the target covariate. Accordingly, the classical ordinary least squares approach can be employed to estimate the regression coefficient associated with the target covariate. In addition, we demonstrate that the resulting estimator is consistent and asymptotically normal even if the random errors are heteroscedastic. This enables us to apply the z-test to assess the significance of each covariate. Based on the p-value obtained from testing the significance of each covariate, we further conduct multiple hypothesis testing by controlling the false discovery rate at the nominal level. Then, we show that the multiple hypothesis testing achieves consistent model selection. Simulation studies and empirical examples are presented to illustrate the finite sample performance and the usefulness of the proposed method, respectively.

AB - In linear regression models with high dimensional data, the classical z-test (or t-test) for testing the significance of each single regression coefficient is no longer applicable. This is mainly because the number of covariates exceeds the sample size. In this paper, we propose a simple and novel alternative by introducing the Correlated Predictors Screening (CPS) method to control for predictors that are highly correlated with the target covariate. Accordingly, the classical ordinary least squares approach can be employed to estimate the regression coefficient associated with the target covariate. In addition, we demonstrate that the resulting estimator is consistent and asymptotically normal even if the random errors are heteroscedastic. This enables us to apply the z-test to assess the significance of each covariate. Based on the p-value obtained from testing the significance of each covariate, we further conduct multiple hypothesis testing by controlling the false discovery rate at the nominal level. Then, we show that the multiple hypothesis testing achieves consistent model selection. Simulation studies and empirical examples are presented to illustrate the finite sample performance and the usefulness of the proposed method, respectively.

UR - http://www.scopus.com/inward/record.url?scp=84990043656&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84990043656&partnerID=8YFLogxK

U2 - 10.1016/j.jeconom.2016.05.016

DO - 10.1016/j.jeconom.2016.05.016

M3 - Article

C2 - 28663668

AN - SCOPUS:84990043656

VL - 195

SP - 154

EP - 168

JO - Journal of Econometrics

JF - Journal of Econometrics

SN - 0304-4076

IS - 1

ER -