Modified efficient importance sampling for partially non-Gaussian state space models

Siem Jan Koopman, Rutger Lit, Thuy Minh Nguyen

Research output: Contribution to JournalArticleAcademicpeer-review

Abstract

The construction of an importance density for partially non-Gaussian state space models is crucial when simulation methods are used for likelihood evaluation, signal extraction, and forecasting. The method of efficient importance sampling is successful in this respect, but we show that it can be implemented in a computationally more efficient manner using standard Kalman filter and smoothing methods. Efficient importance sampling is generally applicable for a wide range of models, but it is typically a custom-built procedure. For the class of partially non-Gaussian state space models, we present a general method for efficient importance sampling. Our novel method makes the efficient importance sampling methodology more accessible because it does not require the computation of a (possibly) complicated density kernel that needs to be tracked for each time period. The new method is illustrated for a stochastic volatility model with a Student's t distribution.

Original languageEnglish
Pages (from-to)44-62
Number of pages19
JournalStatistica Neerlandica
Volume73
Issue number1
DOIs
Publication statusPublished - 1 Feb 2019

Fingerprint

Importance Sampling
State-space Model
Kernel Density
Signal Extraction
Filter Method
Smoothing Methods
Stochastic Volatility Model
t-distribution
Simulation Methods
Kalman Filter
Forecasting
Likelihood
State-space model
Efficient importance sampling
Methodology
Evaluation
Range of data

Keywords

  • efficient importance sampling
  • Kalman filter
  • Monte Carlo maximum likelihood
  • non-Gaussian dynamic models
  • simulation smoothing

Cite this

@article{e341b6e9a3aa4e99b825a74225d19c18,
title = "Modified efficient importance sampling for partially non-Gaussian state space models",
abstract = "The construction of an importance density for partially non-Gaussian state space models is crucial when simulation methods are used for likelihood evaluation, signal extraction, and forecasting. The method of efficient importance sampling is successful in this respect, but we show that it can be implemented in a computationally more efficient manner using standard Kalman filter and smoothing methods. Efficient importance sampling is generally applicable for a wide range of models, but it is typically a custom-built procedure. For the class of partially non-Gaussian state space models, we present a general method for efficient importance sampling. Our novel method makes the efficient importance sampling methodology more accessible because it does not require the computation of a (possibly) complicated density kernel that needs to be tracked for each time period. The new method is illustrated for a stochastic volatility model with a Student's t distribution.",
keywords = "efficient importance sampling, Kalman filter, Monte Carlo maximum likelihood, non-Gaussian dynamic models, simulation smoothing",
author = "Koopman, {Siem Jan} and Rutger Lit and Nguyen, {Thuy Minh}",
year = "2019",
month = "2",
day = "1",
doi = "10.1111/stan.12128",
language = "English",
volume = "73",
pages = "44--62",
journal = "Statistica Neerlandica. Journal of the Netherlands Society for Statistics and Operations Research",
issn = "0039-0402",
publisher = "Wiley-Blackwell",
number = "1",

}

Modified efficient importance sampling for partially non-Gaussian state space models. / Koopman, Siem Jan; Lit, Rutger; Nguyen, Thuy Minh.

In: Statistica Neerlandica, Vol. 73, No. 1, 01.02.2019, p. 44-62.

Research output: Contribution to JournalArticleAcademicpeer-review

TY - JOUR

T1 - Modified efficient importance sampling for partially non-Gaussian state space models

AU - Koopman, Siem Jan

AU - Lit, Rutger

AU - Nguyen, Thuy Minh

PY - 2019/2/1

Y1 - 2019/2/1

N2 - The construction of an importance density for partially non-Gaussian state space models is crucial when simulation methods are used for likelihood evaluation, signal extraction, and forecasting. The method of efficient importance sampling is successful in this respect, but we show that it can be implemented in a computationally more efficient manner using standard Kalman filter and smoothing methods. Efficient importance sampling is generally applicable for a wide range of models, but it is typically a custom-built procedure. For the class of partially non-Gaussian state space models, we present a general method for efficient importance sampling. Our novel method makes the efficient importance sampling methodology more accessible because it does not require the computation of a (possibly) complicated density kernel that needs to be tracked for each time period. The new method is illustrated for a stochastic volatility model with a Student's t distribution.

AB - The construction of an importance density for partially non-Gaussian state space models is crucial when simulation methods are used for likelihood evaluation, signal extraction, and forecasting. The method of efficient importance sampling is successful in this respect, but we show that it can be implemented in a computationally more efficient manner using standard Kalman filter and smoothing methods. Efficient importance sampling is generally applicable for a wide range of models, but it is typically a custom-built procedure. For the class of partially non-Gaussian state space models, we present a general method for efficient importance sampling. Our novel method makes the efficient importance sampling methodology more accessible because it does not require the computation of a (possibly) complicated density kernel that needs to be tracked for each time period. The new method is illustrated for a stochastic volatility model with a Student's t distribution.

KW - efficient importance sampling

KW - Kalman filter

KW - Monte Carlo maximum likelihood

KW - non-Gaussian dynamic models

KW - simulation smoothing

UR - http://www.scopus.com/inward/record.url?scp=85058822110&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85058822110&partnerID=8YFLogxK

U2 - 10.1111/stan.12128

DO - 10.1111/stan.12128

M3 - Article

VL - 73

SP - 44

EP - 62

JO - Statistica Neerlandica. Journal of the Netherlands Society for Statistics and Operations Research

JF - Statistica Neerlandica. Journal of the Netherlands Society for Statistics and Operations Research

SN - 0039-0402

IS - 1

ER -