Bayesian estimation of semiparametric nonlinear dynamic factor analysis models using the Dirichlet process prior

Sy-Miin Chow, Niansheng Tang, Ying Yuan, Xinyuan Song, Hongtu Zhu

Research output: Contribution to journalArticle

27 Citations (Scopus)

Abstract

Parameters in time series and other dynamic models often show complex range restrictions and their distributions may deviate substantially from multivariate normal or other standard parametric distributions. We use the truncated Dirichlet process (DP) as a non-parametric prior for such dynamic parameters in a novel nonlinear Bayesian dynamic factor analysis model. This is equivalent to specifying the prior distribution to be a mixture distribution composed of an unknown number of discrete point masses (or clusters). The stick-breaking prior and the blocked Gibbs sampler are used to enable efficient simulation of posterior samples. Using a series of empirical and simulation examples, we illustrate the flexibility of the proposed approach in approximating distributions of very diverse shapes.

Original languageEnglish (US)
Pages (from-to)69-106
Number of pages38
JournalBritish Journal of Mathematical and Statistical Psychology
Volume64
Issue number1
DOIs
StatePublished - Jan 1 2011

Fingerprint

Dirichlet Process Prior
Nonlinear Dynamics
Bayesian Estimation
Factor Analysis
Dynamic Analysis
Statistical Factor Analysis
Mixture of Distributions
Dirichlet Process
Gibbs Sampler
Multivariate Normal
Prior distribution
Dynamic Model
Simulation
Time series
Flexibility
Model
Restriction
Unknown
Series
Range of data

All Science Journal Classification (ASJC) codes

  • Statistics and Probability
  • Arts and Humanities (miscellaneous)
  • Psychology(all)

Cite this

@article{75d132b397e8467b8bceda8e765f9ad5,
title = "Bayesian estimation of semiparametric nonlinear dynamic factor analysis models using the Dirichlet process prior",
abstract = "Parameters in time series and other dynamic models often show complex range restrictions and their distributions may deviate substantially from multivariate normal or other standard parametric distributions. We use the truncated Dirichlet process (DP) as a non-parametric prior for such dynamic parameters in a novel nonlinear Bayesian dynamic factor analysis model. This is equivalent to specifying the prior distribution to be a mixture distribution composed of an unknown number of discrete point masses (or clusters). The stick-breaking prior and the blocked Gibbs sampler are used to enable efficient simulation of posterior samples. Using a series of empirical and simulation examples, we illustrate the flexibility of the proposed approach in approximating distributions of very diverse shapes.",
author = "Sy-Miin Chow and Niansheng Tang and Ying Yuan and Xinyuan Song and Hongtu Zhu",
year = "2011",
month = "1",
day = "1",
doi = "10.1348/000711010X497262",
language = "English (US)",
volume = "64",
pages = "69--106",
journal = "British Journal of Mathematical and Statistical Psychology",
issn = "0007-1102",
publisher = "Wiley-Blackwell",
number = "1",

}

Bayesian estimation of semiparametric nonlinear dynamic factor analysis models using the Dirichlet process prior. / Chow, Sy-Miin; Tang, Niansheng; Yuan, Ying; Song, Xinyuan; Zhu, Hongtu.

In: British Journal of Mathematical and Statistical Psychology, Vol. 64, No. 1, 01.01.2011, p. 69-106.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Bayesian estimation of semiparametric nonlinear dynamic factor analysis models using the Dirichlet process prior

AU - Chow, Sy-Miin

AU - Tang, Niansheng

AU - Yuan, Ying

AU - Song, Xinyuan

AU - Zhu, Hongtu

PY - 2011/1/1

Y1 - 2011/1/1

N2 - Parameters in time series and other dynamic models often show complex range restrictions and their distributions may deviate substantially from multivariate normal or other standard parametric distributions. We use the truncated Dirichlet process (DP) as a non-parametric prior for such dynamic parameters in a novel nonlinear Bayesian dynamic factor analysis model. This is equivalent to specifying the prior distribution to be a mixture distribution composed of an unknown number of discrete point masses (or clusters). The stick-breaking prior and the blocked Gibbs sampler are used to enable efficient simulation of posterior samples. Using a series of empirical and simulation examples, we illustrate the flexibility of the proposed approach in approximating distributions of very diverse shapes.

AB - Parameters in time series and other dynamic models often show complex range restrictions and their distributions may deviate substantially from multivariate normal or other standard parametric distributions. We use the truncated Dirichlet process (DP) as a non-parametric prior for such dynamic parameters in a novel nonlinear Bayesian dynamic factor analysis model. This is equivalent to specifying the prior distribution to be a mixture distribution composed of an unknown number of discrete point masses (or clusters). The stick-breaking prior and the blocked Gibbs sampler are used to enable efficient simulation of posterior samples. Using a series of empirical and simulation examples, we illustrate the flexibility of the proposed approach in approximating distributions of very diverse shapes.

UR - http://www.scopus.com/inward/record.url?scp=79751487340&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=79751487340&partnerID=8YFLogxK

U2 - 10.1348/000711010X497262

DO - 10.1348/000711010X497262

M3 - Article

C2 - 21506946

AN - SCOPUS:79751487340

VL - 64

SP - 69

EP - 106

JO - British Journal of Mathematical and Statistical Psychology

JF - British Journal of Mathematical and Statistical Psychology

SN - 0007-1102

IS - 1

ER -