Empirical supremum rejection sampling

Brian S Caffo, James G. Booth, A. C. Davison

Research output: Contribution to journalArticle

Abstract

Rejection sampling thins out samples from a candidate density from which it is easy to simulate, to obtain samples from a more awkward target density. A prerequisite is knowledge of the finite supremum of the ratio of the target and candidate densities. This severely restricts application of the method because it can be difficult to calculate the supremum. We use theoretical argument and numerical work to show that a practically perfect sample may be obtained by replacing the exact supremum with the maximum obtained from simulated candidates. We also provide diagnostics for failure of the method caused by a bad choice of candidate distribution. The implication is that essentially no theoretical work is required to apply rejection sampling in many practical cases.

Original languageEnglish (US)
Pages (from-to)745-754
Number of pages10
JournalBiometrika
Volume89
Issue number4
DOIs
StatePublished - 2002

Fingerprint

Rejection Sampling
Supremum
Sampling
Target
sampling
Diagnostics
Calculate
application methods

Keywords

  • Accept-reject
  • Candidate distribution
  • Monte Carlo
  • Sample maximum
  • Super-efficient estimator

ASJC Scopus subject areas

  • Agricultural and Biological Sciences(all)
  • Agricultural and Biological Sciences (miscellaneous)
  • Statistics and Probability
  • Mathematics(all)
  • Applied Mathematics

Cite this

Empirical supremum rejection sampling. / Caffo, Brian S; Booth, James G.; Davison, A. C.

In: Biometrika, Vol. 89, No. 4, 2002, p. 745-754.

Research output: Contribution to journalArticle

Caffo, BS, Booth, JG & Davison, AC 2002, 'Empirical supremum rejection sampling', Biometrika, vol. 89, no. 4, pp. 745-754. https://doi.org/10.1093/biomet/89.4.745
Caffo, Brian S ; Booth, James G. ; Davison, A. C. / Empirical supremum rejection sampling. In: Biometrika. 2002 ; Vol. 89, No. 4. pp. 745-754.
@article{ad44a61f638346e18f4050e6e63d662f,
title = "Empirical supremum rejection sampling",
abstract = "Rejection sampling thins out samples from a candidate density from which it is easy to simulate, to obtain samples from a more awkward target density. A prerequisite is knowledge of the finite supremum of the ratio of the target and candidate densities. This severely restricts application of the method because it can be difficult to calculate the supremum. We use theoretical argument and numerical work to show that a practically perfect sample may be obtained by replacing the exact supremum with the maximum obtained from simulated candidates. We also provide diagnostics for failure of the method caused by a bad choice of candidate distribution. The implication is that essentially no theoretical work is required to apply rejection sampling in many practical cases.",
keywords = "Accept-reject, Candidate distribution, Monte Carlo, Sample maximum, Super-efficient estimator",
author = "Caffo, {Brian S} and Booth, {James G.} and Davison, {A. C.}",
year = "2002",
doi = "10.1093/biomet/89.4.745",
language = "English (US)",
volume = "89",
pages = "745--754",
journal = "Biometrika",
issn = "0006-3444",
publisher = "Oxford University Press",
number = "4",

}

TY - JOUR

T1 - Empirical supremum rejection sampling

AU - Caffo, Brian S

AU - Booth, James G.

AU - Davison, A. C.

PY - 2002

Y1 - 2002

N2 - Rejection sampling thins out samples from a candidate density from which it is easy to simulate, to obtain samples from a more awkward target density. A prerequisite is knowledge of the finite supremum of the ratio of the target and candidate densities. This severely restricts application of the method because it can be difficult to calculate the supremum. We use theoretical argument and numerical work to show that a practically perfect sample may be obtained by replacing the exact supremum with the maximum obtained from simulated candidates. We also provide diagnostics for failure of the method caused by a bad choice of candidate distribution. The implication is that essentially no theoretical work is required to apply rejection sampling in many practical cases.

AB - Rejection sampling thins out samples from a candidate density from which it is easy to simulate, to obtain samples from a more awkward target density. A prerequisite is knowledge of the finite supremum of the ratio of the target and candidate densities. This severely restricts application of the method because it can be difficult to calculate the supremum. We use theoretical argument and numerical work to show that a practically perfect sample may be obtained by replacing the exact supremum with the maximum obtained from simulated candidates. We also provide diagnostics for failure of the method caused by a bad choice of candidate distribution. The implication is that essentially no theoretical work is required to apply rejection sampling in many practical cases.

KW - Accept-reject

KW - Candidate distribution

KW - Monte Carlo

KW - Sample maximum

KW - Super-efficient estimator

UR - http://www.scopus.com/inward/record.url?scp=8644285268&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=8644285268&partnerID=8YFLogxK

U2 - 10.1093/biomet/89.4.745

DO - 10.1093/biomet/89.4.745

M3 - Article

AN - SCOPUS:8644285268

VL - 89

SP - 745

EP - 754

JO - Biometrika

JF - Biometrika

SN - 0006-3444

IS - 4

ER -