How proper are Bayesian models in the astronomical literature?

Hyungsuk Tak, Sujit K. Ghosh, Justin A. Ellis

Research output: Contribution to journalArticle

4 Citations (Scopus)

Abstract

The well-known Bayes theorem assumes that a posterior distribution is a probability distribution. However, the posterior distribution may no longer be a probability distribution if an improper prior distribution (non-probability measure) such as an unbounded uniform prior is used. Improper priors are often used in the astronomical literature to reflect a lack of prior knowledge, but checking whether the resulting posterior is a probability distribution is sometimes neglected. It turns out that 23 out of 75 articles (30.7 per cent) published online in two renowned astronomy journals (ApJ and MNRAS) between 2017 Jan 1 and Oct 15 make use of Bayesian analyses without rigorously establishing posterior propriety. A disturbing aspect is that a Gibbs-type Markov chain Monte Carlo (MCMC) method can produce a seemingly reasonable posterior sample even when the posterior is not a probability distribution (Hobert & Casella 1996). In such cases, researchers may erroneously make probabilistic inferences without noticing that the MCMC sample is from a non-existing probability distribution. We review why checking posterior propriety is fundamental in Bayesian analyses, and discuss how to set up scientifically motivated proper priors.

Original languageEnglish (US)
Pages (from-to)277-285
Number of pages9
JournalMonthly Notices of the Royal Astronomical Society
Volume481
Issue number1
DOIs
StatePublished - Nov 21 2018

Fingerprint

Markov chains
Markov chain
Bayes theorem
distribution
astronomy
inference
Monte Carlo method
method

All Science Journal Classification (ASJC) codes

  • Astronomy and Astrophysics
  • Space and Planetary Science

Cite this

Tak, Hyungsuk ; Ghosh, Sujit K. ; Ellis, Justin A. / How proper are Bayesian models in the astronomical literature?. In: Monthly Notices of the Royal Astronomical Society. 2018 ; Vol. 481, No. 1. pp. 277-285.
@article{02f29ca75dce4935ace36fec35adf6b0,
title = "How proper are Bayesian models in the astronomical literature?",
abstract = "The well-known Bayes theorem assumes that a posterior distribution is a probability distribution. However, the posterior distribution may no longer be a probability distribution if an improper prior distribution (non-probability measure) such as an unbounded uniform prior is used. Improper priors are often used in the astronomical literature to reflect a lack of prior knowledge, but checking whether the resulting posterior is a probability distribution is sometimes neglected. It turns out that 23 out of 75 articles (30.7 per cent) published online in two renowned astronomy journals (ApJ and MNRAS) between 2017 Jan 1 and Oct 15 make use of Bayesian analyses without rigorously establishing posterior propriety. A disturbing aspect is that a Gibbs-type Markov chain Monte Carlo (MCMC) method can produce a seemingly reasonable posterior sample even when the posterior is not a probability distribution (Hobert & Casella 1996). In such cases, researchers may erroneously make probabilistic inferences without noticing that the MCMC sample is from a non-existing probability distribution. We review why checking posterior propriety is fundamental in Bayesian analyses, and discuss how to set up scientifically motivated proper priors.",
author = "Hyungsuk Tak and Ghosh, {Sujit K.} and Ellis, {Justin A.}",
year = "2018",
month = "11",
day = "21",
doi = "10.1093/mnras/sty2326",
language = "English (US)",
volume = "481",
pages = "277--285",
journal = "Monthly Notices of the Royal Astronomical Society",
issn = "0035-8711",
publisher = "Oxford University Press",
number = "1",

}

How proper are Bayesian models in the astronomical literature? / Tak, Hyungsuk; Ghosh, Sujit K.; Ellis, Justin A.

In: Monthly Notices of the Royal Astronomical Society, Vol. 481, No. 1, 21.11.2018, p. 277-285.

Research output: Contribution to journalArticle

TY - JOUR

T1 - How proper are Bayesian models in the astronomical literature?

AU - Tak, Hyungsuk

AU - Ghosh, Sujit K.

AU - Ellis, Justin A.

PY - 2018/11/21

Y1 - 2018/11/21

N2 - The well-known Bayes theorem assumes that a posterior distribution is a probability distribution. However, the posterior distribution may no longer be a probability distribution if an improper prior distribution (non-probability measure) such as an unbounded uniform prior is used. Improper priors are often used in the astronomical literature to reflect a lack of prior knowledge, but checking whether the resulting posterior is a probability distribution is sometimes neglected. It turns out that 23 out of 75 articles (30.7 per cent) published online in two renowned astronomy journals (ApJ and MNRAS) between 2017 Jan 1 and Oct 15 make use of Bayesian analyses without rigorously establishing posterior propriety. A disturbing aspect is that a Gibbs-type Markov chain Monte Carlo (MCMC) method can produce a seemingly reasonable posterior sample even when the posterior is not a probability distribution (Hobert & Casella 1996). In such cases, researchers may erroneously make probabilistic inferences without noticing that the MCMC sample is from a non-existing probability distribution. We review why checking posterior propriety is fundamental in Bayesian analyses, and discuss how to set up scientifically motivated proper priors.

AB - The well-known Bayes theorem assumes that a posterior distribution is a probability distribution. However, the posterior distribution may no longer be a probability distribution if an improper prior distribution (non-probability measure) such as an unbounded uniform prior is used. Improper priors are often used in the astronomical literature to reflect a lack of prior knowledge, but checking whether the resulting posterior is a probability distribution is sometimes neglected. It turns out that 23 out of 75 articles (30.7 per cent) published online in two renowned astronomy journals (ApJ and MNRAS) between 2017 Jan 1 and Oct 15 make use of Bayesian analyses without rigorously establishing posterior propriety. A disturbing aspect is that a Gibbs-type Markov chain Monte Carlo (MCMC) method can produce a seemingly reasonable posterior sample even when the posterior is not a probability distribution (Hobert & Casella 1996). In such cases, researchers may erroneously make probabilistic inferences without noticing that the MCMC sample is from a non-existing probability distribution. We review why checking posterior propriety is fundamental in Bayesian analyses, and discuss how to set up scientifically motivated proper priors.

UR - http://www.scopus.com/inward/record.url?scp=85054084560&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85054084560&partnerID=8YFLogxK

U2 - 10.1093/mnras/sty2326

DO - 10.1093/mnras/sty2326

M3 - Article

AN - SCOPUS:85054084560

VL - 481

SP - 277

EP - 285

JO - Monthly Notices of the Royal Astronomical Society

JF - Monthly Notices of the Royal Astronomical Society

SN - 0035-8711

IS - 1

ER -