Decomposing posterior variance

Paul Gustafson, Bertrand S Clarke

Research output: Contribution to journalArticle

9 Citations (Scopus)

Abstract

We propose a decomposition of posterior variance somewhat in the spirit of an ANOVA decomposition. Terms in this decomposition come in pairs. Given a single parametric model, for instance, one term describes uncertainty arising because the parameter value is unknown while the other describes uncertainty propagated via uncertainty about which prior distribution is appropriate for the parameter. In the context of multiple candidate models and model-averaged estimates, two additional terms emerge resulting in a four-term decomposition. In the context of multiple spaces of models, six terms result. The value of the decomposition is twofold. First, it yields a fuller accounting of uncertainty than methods which condition on data-driven choices of models or model spaces. Second, it constitutes a novel approach to the study of prior influence in Bayesian analysis.

Original languageEnglish (US)
Pages (from-to)311-327
Number of pages17
JournalJournal of Statistical Planning and Inference
Volume119
Issue number2
DOIs
StatePublished - Feb 1 2004

Fingerprint

Decompose
Decomposition
Term
Uncertainty
Multiple Models
Bayesian Analysis
Prior distribution
Parametric Model
Data-driven
Model
Analysis of variance (ANOVA)
Unknown
Estimate
Context

Keywords

  • Bayesian robustness
  • Model averaging
  • Prior sensitivity
  • Standard error

ASJC Scopus subject areas

  • Statistics, Probability and Uncertainty
  • Applied Mathematics
  • Statistics and Probability

Cite this

Decomposing posterior variance. / Gustafson, Paul; Clarke, Bertrand S.

In: Journal of Statistical Planning and Inference, Vol. 119, No. 2, 01.02.2004, p. 311-327.

Research output: Contribution to journalArticle

Gustafson, Paul ; Clarke, Bertrand S. / Decomposing posterior variance. In: Journal of Statistical Planning and Inference. 2004 ; Vol. 119, No. 2. pp. 311-327.
@article{bc8435aa72fd47fb9e2177349b8f8c9e,
title = "Decomposing posterior variance",
abstract = "We propose a decomposition of posterior variance somewhat in the spirit of an ANOVA decomposition. Terms in this decomposition come in pairs. Given a single parametric model, for instance, one term describes uncertainty arising because the parameter value is unknown while the other describes uncertainty propagated via uncertainty about which prior distribution is appropriate for the parameter. In the context of multiple candidate models and model-averaged estimates, two additional terms emerge resulting in a four-term decomposition. In the context of multiple spaces of models, six terms result. The value of the decomposition is twofold. First, it yields a fuller accounting of uncertainty than methods which condition on data-driven choices of models or model spaces. Second, it constitutes a novel approach to the study of prior influence in Bayesian analysis.",
keywords = "Bayesian robustness, Model averaging, Prior sensitivity, Standard error",
author = "Paul Gustafson and Clarke, {Bertrand S}",
year = "2004",
month = "2",
day = "1",
doi = "10.1016/S0378-3758(02)00491-3",
language = "English (US)",
volume = "119",
pages = "311--327",
journal = "Journal of Statistical Planning and Inference",
issn = "0378-3758",
publisher = "Elsevier",
number = "2",

}

TY - JOUR

T1 - Decomposing posterior variance

AU - Gustafson, Paul

AU - Clarke, Bertrand S

PY - 2004/2/1

Y1 - 2004/2/1

N2 - We propose a decomposition of posterior variance somewhat in the spirit of an ANOVA decomposition. Terms in this decomposition come in pairs. Given a single parametric model, for instance, one term describes uncertainty arising because the parameter value is unknown while the other describes uncertainty propagated via uncertainty about which prior distribution is appropriate for the parameter. In the context of multiple candidate models and model-averaged estimates, two additional terms emerge resulting in a four-term decomposition. In the context of multiple spaces of models, six terms result. The value of the decomposition is twofold. First, it yields a fuller accounting of uncertainty than methods which condition on data-driven choices of models or model spaces. Second, it constitutes a novel approach to the study of prior influence in Bayesian analysis.

AB - We propose a decomposition of posterior variance somewhat in the spirit of an ANOVA decomposition. Terms in this decomposition come in pairs. Given a single parametric model, for instance, one term describes uncertainty arising because the parameter value is unknown while the other describes uncertainty propagated via uncertainty about which prior distribution is appropriate for the parameter. In the context of multiple candidate models and model-averaged estimates, two additional terms emerge resulting in a four-term decomposition. In the context of multiple spaces of models, six terms result. The value of the decomposition is twofold. First, it yields a fuller accounting of uncertainty than methods which condition on data-driven choices of models or model spaces. Second, it constitutes a novel approach to the study of prior influence in Bayesian analysis.

KW - Bayesian robustness

KW - Model averaging

KW - Prior sensitivity

KW - Standard error

UR - http://www.scopus.com/inward/record.url?scp=0242551036&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0242551036&partnerID=8YFLogxK

U2 - 10.1016/S0378-3758(02)00491-3

DO - 10.1016/S0378-3758(02)00491-3

M3 - Article

VL - 119

SP - 311

EP - 327

JO - Journal of Statistical Planning and Inference

JF - Journal of Statistical Planning and Inference

SN - 0378-3758

IS - 2

ER -