Error bounds of MCMC for functions with unbounded stationary variance

Daniel Rudolf*, Nikolaus Schweizer

*Corresponding author for this work

Research output: Contribution to journalArticleScientificpeer-review

Abstract

We prove explicit error bounds for Markov chain Monte Carlo (MCMC) methods to compute expectations of functions with unbounded stationary variance. We assume that there is a p is an element of (1, 2) so that the functions have finite L-p-norm. For uniformly ergodic Markov chains we obtain error bounds with the optimal order of convergence n(1/p-1) and if there exists a spectral gap we almost get the optimal order. Further, a burn-in period is taken into account and a recipe for choosing the burn-in is provided. (C) 2015 Elsevier B.V. All rights reserved.

Original languageEnglish
Pages (from-to)6-12
JournalStatistics & probability letters
Volume99
DOIs
Publication statusPublished - Apr 2015
Externally publishedYes

Keywords

  • Markov chain Monte Carlo
  • Absolute mean error
  • Uniform ergodicity
  • Spectral gap

Cite this

@article{a902e00e1b8f4b958cdffc4206f63b71,
title = "Error bounds of MCMC for functions with unbounded stationary variance",
abstract = "We prove explicit error bounds for Markov chain Monte Carlo (MCMC) methods to compute expectations of functions with unbounded stationary variance. We assume that there is a p is an element of (1, 2) so that the functions have finite L-p-norm. For uniformly ergodic Markov chains we obtain error bounds with the optimal order of convergence n(1/p-1) and if there exists a spectral gap we almost get the optimal order. Further, a burn-in period is taken into account and a recipe for choosing the burn-in is provided. (C) 2015 Elsevier B.V. All rights reserved.",
keywords = "Markov chain Monte Carlo, Absolute mean error, Uniform ergodicity, Spectral gap",
author = "Daniel Rudolf and Nikolaus Schweizer",
year = "2015",
month = "4",
doi = "10.1016/j.spl.2014.07.035",
language = "English",
volume = "99",
pages = "6--12",
journal = "Statistics & probability letters",
issn = "0167-7152",
publisher = "Elsevier Science BV",

}

Error bounds of MCMC for functions with unbounded stationary variance. / Rudolf, Daniel; Schweizer, Nikolaus.

In: Statistics & probability letters, Vol. 99, 04.2015, p. 6-12.

Research output: Contribution to journalArticleScientificpeer-review

TY - JOUR

T1 - Error bounds of MCMC for functions with unbounded stationary variance

AU - Rudolf, Daniel

AU - Schweizer, Nikolaus

PY - 2015/4

Y1 - 2015/4

N2 - We prove explicit error bounds for Markov chain Monte Carlo (MCMC) methods to compute expectations of functions with unbounded stationary variance. We assume that there is a p is an element of (1, 2) so that the functions have finite L-p-norm. For uniformly ergodic Markov chains we obtain error bounds with the optimal order of convergence n(1/p-1) and if there exists a spectral gap we almost get the optimal order. Further, a burn-in period is taken into account and a recipe for choosing the burn-in is provided. (C) 2015 Elsevier B.V. All rights reserved.

AB - We prove explicit error bounds for Markov chain Monte Carlo (MCMC) methods to compute expectations of functions with unbounded stationary variance. We assume that there is a p is an element of (1, 2) so that the functions have finite L-p-norm. For uniformly ergodic Markov chains we obtain error bounds with the optimal order of convergence n(1/p-1) and if there exists a spectral gap we almost get the optimal order. Further, a burn-in period is taken into account and a recipe for choosing the burn-in is provided. (C) 2015 Elsevier B.V. All rights reserved.

KW - Markov chain Monte Carlo

KW - Absolute mean error

KW - Uniform ergodicity

KW - Spectral gap

U2 - 10.1016/j.spl.2014.07.035

DO - 10.1016/j.spl.2014.07.035

M3 - Article

VL - 99

SP - 6

EP - 12

JO - Statistics & probability letters

JF - Statistics & probability letters

SN - 0167-7152

ER -