Bayesian high-dimensional regression for change point analysis

Abhirup Datta, Hui Zou, Sudipto Banerjee

Research output: Contribution to journalArticle

Abstract

In many econometrics applications, the dataset under investigation spans heterogeneous regimes that are more appropriately modeled using piece-wise components for each of the data segments separated by change-points. We consider using Bayesian high-dimensional shrinkage priors in a change point setting to understand segment-specific relationship between the response and the covariates. Covariate selection before and after each change point can identify possibly different sets of relevant covariates, while the fully Bayesian approach ensures posterior inference for the change points is also available. We demonstrate the flexibility of the approach for imposing different variable selection constraints like grouping or partial selection and discuss strategies to detect an unknown number of change points. Simulation experiments reveal that this simple approach delivers accurate variable selection, and inference on location of the change points, and substantially outperforms a frequentist lasso-based approach, uniformly across a wide range of scenarios. Application of our model to Minnesota house price dataset reveals change in the relationship between house and stock prices around the sub-prime mortgage crisis.

Original languageEnglish (US)
Pages (from-to)253-264
Number of pages12
JournalStatistics and its Interface
Volume12
Issue number2
DOIs
StatePublished - Jan 1 2019

Fingerprint

Change-point Analysis
Change Point
High-dimensional
Regression
Covariates
Variable Selection
Lasso
Experiments
Stock Prices
Shrinkage
Econometrics
Bayesian Approach
Grouping
Simulation Experiment
Flexibility
Partial
Unknown
Scenarios
Range of data
Demonstrate

Keywords

  • Bayesian inference
  • Change point detection
  • High-dimensional regression
  • Markov Chain Monte Carlo
  • Minnesota House Price Data
  • Variable selection

PubMed: MeSH publication types

  • Journal Article

Cite this

Bayesian high-dimensional regression for change point analysis. / Datta, Abhirup; Zou, Hui; Banerjee, Sudipto.

In: Statistics and its Interface, Vol. 12, No. 2, 01.01.2019, p. 253-264.

Research output: Contribution to journalArticle

Datta, Abhirup ; Zou, Hui ; Banerjee, Sudipto. / Bayesian high-dimensional regression for change point analysis. In: Statistics and its Interface. 2019 ; Vol. 12, No. 2. pp. 253-264.
@article{91c9ca270c254bddb10dccec6263f193,
title = "Bayesian high-dimensional regression for change point analysis",
abstract = "In many econometrics applications, the dataset under investigation spans heterogeneous regimes that are more appropriately modeled using piece-wise components for each of the data segments separated by change-points. We consider using Bayesian high-dimensional shrinkage priors in a change point setting to understand segment-specific relationship between the response and the covariates. Covariate selection before and after each change point can identify possibly different sets of relevant covariates, while the fully Bayesian approach ensures posterior inference for the change points is also available. We demonstrate the flexibility of the approach for imposing different variable selection constraints like grouping or partial selection and discuss strategies to detect an unknown number of change points. Simulation experiments reveal that this simple approach delivers accurate variable selection, and inference on location of the change points, and substantially outperforms a frequentist lasso-based approach, uniformly across a wide range of scenarios. Application of our model to Minnesota house price dataset reveals change in the relationship between house and stock prices around the sub-prime mortgage crisis.",
keywords = "Bayesian inference, Change point detection, High-dimensional regression, Markov Chain Monte Carlo, Minnesota House Price Data, Variable selection",
author = "Abhirup Datta and Hui Zou and Sudipto Banerjee",
year = "2019",
month = "1",
day = "1",
doi = "10.4310/SII.2019.v12.n2.a6",
language = "English (US)",
volume = "12",
pages = "253--264",
journal = "Statistics and its Interface",
issn = "1938-7989",
publisher = "International Press of Boston, Inc.",
number = "2",

}

TY - JOUR

T1 - Bayesian high-dimensional regression for change point analysis

AU - Datta, Abhirup

AU - Zou, Hui

AU - Banerjee, Sudipto

PY - 2019/1/1

Y1 - 2019/1/1

N2 - In many econometrics applications, the dataset under investigation spans heterogeneous regimes that are more appropriately modeled using piece-wise components for each of the data segments separated by change-points. We consider using Bayesian high-dimensional shrinkage priors in a change point setting to understand segment-specific relationship between the response and the covariates. Covariate selection before and after each change point can identify possibly different sets of relevant covariates, while the fully Bayesian approach ensures posterior inference for the change points is also available. We demonstrate the flexibility of the approach for imposing different variable selection constraints like grouping or partial selection and discuss strategies to detect an unknown number of change points. Simulation experiments reveal that this simple approach delivers accurate variable selection, and inference on location of the change points, and substantially outperforms a frequentist lasso-based approach, uniformly across a wide range of scenarios. Application of our model to Minnesota house price dataset reveals change in the relationship between house and stock prices around the sub-prime mortgage crisis.

AB - In many econometrics applications, the dataset under investigation spans heterogeneous regimes that are more appropriately modeled using piece-wise components for each of the data segments separated by change-points. We consider using Bayesian high-dimensional shrinkage priors in a change point setting to understand segment-specific relationship between the response and the covariates. Covariate selection before and after each change point can identify possibly different sets of relevant covariates, while the fully Bayesian approach ensures posterior inference for the change points is also available. We demonstrate the flexibility of the approach for imposing different variable selection constraints like grouping or partial selection and discuss strategies to detect an unknown number of change points. Simulation experiments reveal that this simple approach delivers accurate variable selection, and inference on location of the change points, and substantially outperforms a frequentist lasso-based approach, uniformly across a wide range of scenarios. Application of our model to Minnesota house price dataset reveals change in the relationship between house and stock prices around the sub-prime mortgage crisis.

KW - Bayesian inference

KW - Change point detection

KW - High-dimensional regression

KW - Markov Chain Monte Carlo

KW - Minnesota House Price Data

KW - Variable selection

UR - http://www.scopus.com/inward/record.url?scp=85063726826&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85063726826&partnerID=8YFLogxK

U2 - 10.4310/SII.2019.v12.n2.a6

DO - 10.4310/SII.2019.v12.n2.a6

M3 - Article

C2 - 31543930

AN - SCOPUS:85063726826

VL - 12

SP - 253

EP - 264

JO - Statistics and its Interface

JF - Statistics and its Interface

SN - 1938-7989

IS - 2

ER -