Auxiliary Likelihood-Based Approximate Bayesian Computation in State Space Models

Gael M. Martin, Brendan P.M. McCabe, David T. Frazier, Worapree Maneesoonthorn, Christian P. Robert

Research output: Contribution to journalArticleResearchpeer-review

Abstract

A computationally simple approach to inference in state space models is proposed, using approximate Bayesian computation (ABC). ABC avoids evaluation of an intractable likelihood by matching summary statistics for the observed data with statistics computed from data simulated from the true process, based on parameter draws from the prior. Draws that produce a “match” between observed and simulated summaries are retained, and used to estimate the inaccessible posterior. With no reduction to a low-dimensional set ofsufficient statistics being possible in the state space setting, we define the summaries as the maximum of an auxiliary likelihood function, and thereby exploit the asymptotic sufficiency of this estimator for the auxiliary parameter vector. We derive conditions under which this approach—including a computationally efficient version based on the auxiliary score—achieves Bayesian consistency. To reduce the well-documented inaccuracy of ABC in multiparameter settings, we propose the separate treatment of each parameter dimension using an integrated likelihood technique. Three stochastic volatility models for which exact Bayesian inference is either computationally challenging, or infeasible, are used for illustration. We demonstrate that our approach compares favorably against an extensive set of approximate and exact comparators. An empirical illustration completes the article. Supplementary materials for this article are available online.

Original languageEnglish
Number of pages34
JournalJournal of Computational and Graphical Statistics
DOIs
Publication statusAccepted/In press - 18 Dec 2018

Keywords

  • Alpha-stable distribution
  • Asymptotic sufficiency
  • Bayesian consistency
  • Likelihood-free method
  • Stochastic volatility model
  • Unscented Kalman filter

Cite this

@article{31f7ecd5a7ee4cbabaf0907608341fd2,
title = "Auxiliary Likelihood-Based Approximate Bayesian Computation in State Space Models",
abstract = "A computationally simple approach to inference in state space models is proposed, using approximate Bayesian computation (ABC). ABC avoids evaluation of an intractable likelihood by matching summary statistics for the observed data with statistics computed from data simulated from the true process, based on parameter draws from the prior. Draws that produce a “match” between observed and simulated summaries are retained, and used to estimate the inaccessible posterior. With no reduction to a low-dimensional set ofsufficient statistics being possible in the state space setting, we define the summaries as the maximum of an auxiliary likelihood function, and thereby exploit the asymptotic sufficiency of this estimator for the auxiliary parameter vector. We derive conditions under which this approach—including a computationally efficient version based on the auxiliary score—achieves Bayesian consistency. To reduce the well-documented inaccuracy of ABC in multiparameter settings, we propose the separate treatment of each parameter dimension using an integrated likelihood technique. Three stochastic volatility models for which exact Bayesian inference is either computationally challenging, or infeasible, are used for illustration. We demonstrate that our approach compares favorably against an extensive set of approximate and exact comparators. An empirical illustration completes the article. Supplementary materials for this article are available online.",
keywords = "Alpha-stable distribution, Asymptotic sufficiency, Bayesian consistency, Likelihood-free method, Stochastic volatility model, Unscented Kalman filter",
author = "Martin, {Gael M.} and McCabe, {Brendan P.M.} and Frazier, {David T.} and Worapree Maneesoonthorn and Robert, {Christian P.}",
year = "2018",
month = "12",
day = "18",
doi = "10.1080/10618600.2018.1552154",
language = "English",
journal = "Journal of Computational and Graphical Statistics",
issn = "1061-8600",
publisher = "Taylor & Francis",

}

Auxiliary Likelihood-Based Approximate Bayesian Computation in State Space Models. / Martin, Gael M.; McCabe, Brendan P.M.; Frazier, David T.; Maneesoonthorn, Worapree; Robert, Christian P.

In: Journal of Computational and Graphical Statistics, 18.12.2018.

Research output: Contribution to journalArticleResearchpeer-review

TY - JOUR

T1 - Auxiliary Likelihood-Based Approximate Bayesian Computation in State Space Models

AU - Martin, Gael M.

AU - McCabe, Brendan P.M.

AU - Frazier, David T.

AU - Maneesoonthorn, Worapree

AU - Robert, Christian P.

PY - 2018/12/18

Y1 - 2018/12/18

N2 - A computationally simple approach to inference in state space models is proposed, using approximate Bayesian computation (ABC). ABC avoids evaluation of an intractable likelihood by matching summary statistics for the observed data with statistics computed from data simulated from the true process, based on parameter draws from the prior. Draws that produce a “match” between observed and simulated summaries are retained, and used to estimate the inaccessible posterior. With no reduction to a low-dimensional set ofsufficient statistics being possible in the state space setting, we define the summaries as the maximum of an auxiliary likelihood function, and thereby exploit the asymptotic sufficiency of this estimator for the auxiliary parameter vector. We derive conditions under which this approach—including a computationally efficient version based on the auxiliary score—achieves Bayesian consistency. To reduce the well-documented inaccuracy of ABC in multiparameter settings, we propose the separate treatment of each parameter dimension using an integrated likelihood technique. Three stochastic volatility models for which exact Bayesian inference is either computationally challenging, or infeasible, are used for illustration. We demonstrate that our approach compares favorably against an extensive set of approximate and exact comparators. An empirical illustration completes the article. Supplementary materials for this article are available online.

AB - A computationally simple approach to inference in state space models is proposed, using approximate Bayesian computation (ABC). ABC avoids evaluation of an intractable likelihood by matching summary statistics for the observed data with statistics computed from data simulated from the true process, based on parameter draws from the prior. Draws that produce a “match” between observed and simulated summaries are retained, and used to estimate the inaccessible posterior. With no reduction to a low-dimensional set ofsufficient statistics being possible in the state space setting, we define the summaries as the maximum of an auxiliary likelihood function, and thereby exploit the asymptotic sufficiency of this estimator for the auxiliary parameter vector. We derive conditions under which this approach—including a computationally efficient version based on the auxiliary score—achieves Bayesian consistency. To reduce the well-documented inaccuracy of ABC in multiparameter settings, we propose the separate treatment of each parameter dimension using an integrated likelihood technique. Three stochastic volatility models for which exact Bayesian inference is either computationally challenging, or infeasible, are used for illustration. We demonstrate that our approach compares favorably against an extensive set of approximate and exact comparators. An empirical illustration completes the article. Supplementary materials for this article are available online.

KW - Alpha-stable distribution

KW - Asymptotic sufficiency

KW - Bayesian consistency

KW - Likelihood-free method

KW - Stochastic volatility model

KW - Unscented Kalman filter

UR - http://www.scopus.com/inward/record.url?scp=85062349437&partnerID=8YFLogxK

U2 - 10.1080/10618600.2018.1552154

DO - 10.1080/10618600.2018.1552154

M3 - Article

JO - Journal of Computational and Graphical Statistics

JF - Journal of Computational and Graphical Statistics

SN - 1061-8600

ER -