Approximate Bayesian forecasting

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

Research output: Contribution to journalArticleResearchpeer-review

1 Citation (Scopus)

Abstract

Approximate Bayesian Computation (ABC) has become increasingly prominent as a method for conducting parameter inference in a range of challenging statistical problems, most notably those characterized by an intractable likelihood function. In this paper, we focus on the use of ABC not as a tool for parametric inference, but as a means of generating probabilistic forecasts; or for conducting what we refer to as ‘approximate Bayesian forecasting’. The four key issues explored are: (i) the link between the theoretical behavior of the ABC posterior and that of the ABC-based predictive; (ii) the use of proper scoring rules to measure the (potential) loss of forecast accuracy when using an approximate rather than an exact predictive; (iii) the performance of approximate Bayesian forecasting in state space models; and (iv) the use of forecasting criteria to inform the selection of ABC summaries in empirical settings. The primary finding of the paper is that ABC can provide a computationally efficient means of generating probabilistic forecasts that are nearly identical to those produced by the exact predictive, and in a fraction of the time required to produce predictions via an exact method.

Original languageEnglish
Pages (from-to)521-539
Number of pages19
JournalInternational Journal of Forecasting
Volume35
Issue number2
DOIs
Publication statusPublished - 1 Apr 2019

Keywords

  • Bayesian prediction
  • Jump-diffusion models
  • Likelihood-free methods
  • Particle filtering
  • Predictive merging
  • Proper scoring rules

Cite this

Frazier, David T. ; Maneesoonthorn, Worapree ; Martin, Gael M. ; McCabe, Brendan P.M. / Approximate Bayesian forecasting. In: International Journal of Forecasting. 2019 ; Vol. 35, No. 2. pp. 521-539.
@article{bfca0ba33f3f42e8955ccb6c80852a01,
title = "Approximate Bayesian forecasting",
abstract = "Approximate Bayesian Computation (ABC) has become increasingly prominent as a method for conducting parameter inference in a range of challenging statistical problems, most notably those characterized by an intractable likelihood function. In this paper, we focus on the use of ABC not as a tool for parametric inference, but as a means of generating probabilistic forecasts; or for conducting what we refer to as ‘approximate Bayesian forecasting’. The four key issues explored are: (i) the link between the theoretical behavior of the ABC posterior and that of the ABC-based predictive; (ii) the use of proper scoring rules to measure the (potential) loss of forecast accuracy when using an approximate rather than an exact predictive; (iii) the performance of approximate Bayesian forecasting in state space models; and (iv) the use of forecasting criteria to inform the selection of ABC summaries in empirical settings. The primary finding of the paper is that ABC can provide a computationally efficient means of generating probabilistic forecasts that are nearly identical to those produced by the exact predictive, and in a fraction of the time required to produce predictions via an exact method.",
keywords = "Bayesian prediction, Jump-diffusion models, Likelihood-free methods, Particle filtering, Predictive merging, Proper scoring rules",
author = "Frazier, {David T.} and Worapree Maneesoonthorn and Martin, {Gael M.} and McCabe, {Brendan P.M.}",
year = "2019",
month = "4",
day = "1",
doi = "10.1016/j.ijforecast.2018.08.003",
language = "English",
volume = "35",
pages = "521--539",
journal = "International Journal of Forecasting",
issn = "0169-2070",
publisher = "Elsevier",
number = "2",

}

Approximate Bayesian forecasting. / Frazier, David T.; Maneesoonthorn, Worapree; Martin, Gael M.; McCabe, Brendan P.M.

In: International Journal of Forecasting, Vol. 35, No. 2, 01.04.2019, p. 521-539.

Research output: Contribution to journalArticleResearchpeer-review

TY - JOUR

T1 - Approximate Bayesian forecasting

AU - Frazier, David T.

AU - Maneesoonthorn, Worapree

AU - Martin, Gael M.

AU - McCabe, Brendan P.M.

PY - 2019/4/1

Y1 - 2019/4/1

N2 - Approximate Bayesian Computation (ABC) has become increasingly prominent as a method for conducting parameter inference in a range of challenging statistical problems, most notably those characterized by an intractable likelihood function. In this paper, we focus on the use of ABC not as a tool for parametric inference, but as a means of generating probabilistic forecasts; or for conducting what we refer to as ‘approximate Bayesian forecasting’. The four key issues explored are: (i) the link between the theoretical behavior of the ABC posterior and that of the ABC-based predictive; (ii) the use of proper scoring rules to measure the (potential) loss of forecast accuracy when using an approximate rather than an exact predictive; (iii) the performance of approximate Bayesian forecasting in state space models; and (iv) the use of forecasting criteria to inform the selection of ABC summaries in empirical settings. The primary finding of the paper is that ABC can provide a computationally efficient means of generating probabilistic forecasts that are nearly identical to those produced by the exact predictive, and in a fraction of the time required to produce predictions via an exact method.

AB - Approximate Bayesian Computation (ABC) has become increasingly prominent as a method for conducting parameter inference in a range of challenging statistical problems, most notably those characterized by an intractable likelihood function. In this paper, we focus on the use of ABC not as a tool for parametric inference, but as a means of generating probabilistic forecasts; or for conducting what we refer to as ‘approximate Bayesian forecasting’. The four key issues explored are: (i) the link between the theoretical behavior of the ABC posterior and that of the ABC-based predictive; (ii) the use of proper scoring rules to measure the (potential) loss of forecast accuracy when using an approximate rather than an exact predictive; (iii) the performance of approximate Bayesian forecasting in state space models; and (iv) the use of forecasting criteria to inform the selection of ABC summaries in empirical settings. The primary finding of the paper is that ABC can provide a computationally efficient means of generating probabilistic forecasts that are nearly identical to those produced by the exact predictive, and in a fraction of the time required to produce predictions via an exact method.

KW - Bayesian prediction

KW - Jump-diffusion models

KW - Likelihood-free methods

KW - Particle filtering

KW - Predictive merging

KW - Proper scoring rules

UR - http://www.scopus.com/inward/record.url?scp=85059463199&partnerID=8YFLogxK

U2 - 10.1016/j.ijforecast.2018.08.003

DO - 10.1016/j.ijforecast.2018.08.003

M3 - Article

VL - 35

SP - 521

EP - 539

JO - International Journal of Forecasting

JF - International Journal of Forecasting

SN - 0169-2070

IS - 2

ER -