Partial factor modeling: Predictor-dependent shrinkage for linear regression

Paul Hahn, Carlos M. Carvalho, Sayan Mukherjee

Research output: Contribution to journalArticle

5 Citations (Scopus)

Abstract

We develop a modified Gaussian factor model for the purpose of inducing predictor-dependent shrinkage for linear regression. The new model predicts well across a wide range of covariance structures, on real and simulated data. Furthermore, the new model facilitates variable selection in the case of correlated predictor variables, which often stymies other methods.

Original languageEnglish (US)
Pages (from-to)999-1008
Number of pages10
JournalJournal of the American Statistical Association
Volume108
Issue number503
DOIs
StatePublished - Dec 16 2013
Externally publishedYes

Fingerprint

Shrinkage
Linear regression
Predictors
Partial
Dependent
Factor Models
Covariance Structure
Gaussian Model
Variable Selection
Modeling
Model Selection
Predict
Range of data
Factors
Model
Variable selection

Keywords

  • g Prior
  • Prediction
  • Shrinkage estimators
  • Variable selection

ASJC Scopus subject areas

  • Statistics and Probability
  • Statistics, Probability and Uncertainty

Cite this

Partial factor modeling : Predictor-dependent shrinkage for linear regression. / Hahn, Paul; Carvalho, Carlos M.; Mukherjee, Sayan.

In: Journal of the American Statistical Association, Vol. 108, No. 503, 16.12.2013, p. 999-1008.

Research output: Contribution to journalArticle

Hahn, Paul ; Carvalho, Carlos M. ; Mukherjee, Sayan. / Partial factor modeling : Predictor-dependent shrinkage for linear regression. In: Journal of the American Statistical Association. 2013 ; Vol. 108, No. 503. pp. 999-1008.
@article{b24b4935496142bcadf4825f8bffa4fb,
title = "Partial factor modeling: Predictor-dependent shrinkage for linear regression",
abstract = "We develop a modified Gaussian factor model for the purpose of inducing predictor-dependent shrinkage for linear regression. The new model predicts well across a wide range of covariance structures, on real and simulated data. Furthermore, the new model facilitates variable selection in the case of correlated predictor variables, which often stymies other methods.",
keywords = "g Prior, Prediction, Shrinkage estimators, Variable selection",
author = "Paul Hahn and Carvalho, {Carlos M.} and Sayan Mukherjee",
year = "2013",
month = "12",
day = "16",
doi = "10.1080/01621459.2013.779843",
language = "English (US)",
volume = "108",
pages = "999--1008",
journal = "Journal of the American Statistical Association",
issn = "0162-1459",
publisher = "Taylor and Francis Ltd.",
number = "503",

}

TY - JOUR

T1 - Partial factor modeling

T2 - Predictor-dependent shrinkage for linear regression

AU - Hahn, Paul

AU - Carvalho, Carlos M.

AU - Mukherjee, Sayan

PY - 2013/12/16

Y1 - 2013/12/16

N2 - We develop a modified Gaussian factor model for the purpose of inducing predictor-dependent shrinkage for linear regression. The new model predicts well across a wide range of covariance structures, on real and simulated data. Furthermore, the new model facilitates variable selection in the case of correlated predictor variables, which often stymies other methods.

AB - We develop a modified Gaussian factor model for the purpose of inducing predictor-dependent shrinkage for linear regression. The new model predicts well across a wide range of covariance structures, on real and simulated data. Furthermore, the new model facilitates variable selection in the case of correlated predictor variables, which often stymies other methods.

KW - g Prior

KW - Prediction

KW - Shrinkage estimators

KW - Variable selection

UR - http://www.scopus.com/inward/record.url?scp=84890013712&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84890013712&partnerID=8YFLogxK

U2 - 10.1080/01621459.2013.779843

DO - 10.1080/01621459.2013.779843

M3 - Article

AN - SCOPUS:84890013712

VL - 108

SP - 999

EP - 1008

JO - Journal of the American Statistical Association

JF - Journal of the American Statistical Association

SN - 0162-1459

IS - 503

ER -