MEBoost: Variable selection in the presence of measurement error

Ben Brown, Timothy Weaver, Julian Wolfson

Research output: Contribution to journalArticle

Abstract

We present a novel method for variable selection in regression models when covariates are measured with error. The iterative algorithm we propose, Measurement Error Boosting (MEBoost), follows a path defined by estimating equations that correct for covariate measurement error. We illustrate the use of MEBoost in practice by analyzing data from the Box Lunch Study, a clinical trial in nutrition where several variables are based on self-report and, hence, measured with error, where we are interested in performing model selection from a large data set to select variables that are related to the number of times a subject binge ate in the last 28 days. Furthermore, we evaluated our method and compared its performance to the recently proposed Convex Conditioned Lasso and to the “naive” Lasso, which does not correct for measurement error through a simulation study. Increasing the degree of measurement error increased prediction error and decreased the probability of accurate covariate selection, but this loss of accuracy occurred to a lesser degree when using MEBoost. Through simulations, we also make a case for the consistency of the model selected.

Original languageEnglish (US)
Pages (from-to)2705-2718
Number of pages14
JournalStatistics in Medicine
Volume38
Issue number15
DOIs
StatePublished - Jul 10 2019

Fingerprint

Boosting
Variable Selection
Measurement Error
Lunch
Self Report
Covariates
Lasso
Clinical Trials
Nutrition
Estimating Equation
Several Variables
Prediction Error
Large Data Sets
Model Selection
Iterative Algorithm
alachlor
Regression Model
Simulation Study
Path
Datasets

Keywords

  • boosting
  • high-dimensional data
  • machine learning
  • measurement error
  • variable selection

Cite this

MEBoost : Variable selection in the presence of measurement error. / Brown, Ben; Weaver, Timothy; Wolfson, Julian.

In: Statistics in Medicine, Vol. 38, No. 15, 10.07.2019, p. 2705-2718.

Research output: Contribution to journalArticle

Brown, Ben ; Weaver, Timothy ; Wolfson, Julian. / MEBoost : Variable selection in the presence of measurement error. In: Statistics in Medicine. 2019 ; Vol. 38, No. 15. pp. 2705-2718.
@article{bf8c63e76e4a4ae1bcfaaf7f89e362aa,
title = "MEBoost: Variable selection in the presence of measurement error",
abstract = "We present a novel method for variable selection in regression models when covariates are measured with error. The iterative algorithm we propose, Measurement Error Boosting (MEBoost), follows a path defined by estimating equations that correct for covariate measurement error. We illustrate the use of MEBoost in practice by analyzing data from the Box Lunch Study, a clinical trial in nutrition where several variables are based on self-report and, hence, measured with error, where we are interested in performing model selection from a large data set to select variables that are related to the number of times a subject binge ate in the last 28 days. Furthermore, we evaluated our method and compared its performance to the recently proposed Convex Conditioned Lasso and to the “naive” Lasso, which does not correct for measurement error through a simulation study. Increasing the degree of measurement error increased prediction error and decreased the probability of accurate covariate selection, but this loss of accuracy occurred to a lesser degree when using MEBoost. Through simulations, we also make a case for the consistency of the model selected.",
keywords = "boosting, high-dimensional data, machine learning, measurement error, variable selection",
author = "Ben Brown and Timothy Weaver and Julian Wolfson",
year = "2019",
month = "7",
day = "10",
doi = "10.1002/sim.8130",
language = "English (US)",
volume = "38",
pages = "2705--2718",
journal = "Statistics in Medicine",
issn = "0277-6715",
publisher = "John Wiley and Sons Ltd",
number = "15",

}

TY - JOUR

T1 - MEBoost

T2 - Variable selection in the presence of measurement error

AU - Brown, Ben

AU - Weaver, Timothy

AU - Wolfson, Julian

PY - 2019/7/10

Y1 - 2019/7/10

N2 - We present a novel method for variable selection in regression models when covariates are measured with error. The iterative algorithm we propose, Measurement Error Boosting (MEBoost), follows a path defined by estimating equations that correct for covariate measurement error. We illustrate the use of MEBoost in practice by analyzing data from the Box Lunch Study, a clinical trial in nutrition where several variables are based on self-report and, hence, measured with error, where we are interested in performing model selection from a large data set to select variables that are related to the number of times a subject binge ate in the last 28 days. Furthermore, we evaluated our method and compared its performance to the recently proposed Convex Conditioned Lasso and to the “naive” Lasso, which does not correct for measurement error through a simulation study. Increasing the degree of measurement error increased prediction error and decreased the probability of accurate covariate selection, but this loss of accuracy occurred to a lesser degree when using MEBoost. Through simulations, we also make a case for the consistency of the model selected.

AB - We present a novel method for variable selection in regression models when covariates are measured with error. The iterative algorithm we propose, Measurement Error Boosting (MEBoost), follows a path defined by estimating equations that correct for covariate measurement error. We illustrate the use of MEBoost in practice by analyzing data from the Box Lunch Study, a clinical trial in nutrition where several variables are based on self-report and, hence, measured with error, where we are interested in performing model selection from a large data set to select variables that are related to the number of times a subject binge ate in the last 28 days. Furthermore, we evaluated our method and compared its performance to the recently proposed Convex Conditioned Lasso and to the “naive” Lasso, which does not correct for measurement error through a simulation study. Increasing the degree of measurement error increased prediction error and decreased the probability of accurate covariate selection, but this loss of accuracy occurred to a lesser degree when using MEBoost. Through simulations, we also make a case for the consistency of the model selected.

KW - boosting

KW - high-dimensional data

KW - machine learning

KW - measurement error

KW - variable selection

UR - http://www.scopus.com/inward/record.url?scp=85062785118&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85062785118&partnerID=8YFLogxK

U2 - 10.1002/sim.8130

DO - 10.1002/sim.8130

M3 - Article

VL - 38

SP - 2705

EP - 2718

JO - Statistics in Medicine

JF - Statistics in Medicine

SN - 0277-6715

IS - 15

ER -