samplers: Particle Filtering MCMC Sampling Algorithms

Description Usage Arguments RW_PF sampler RW_PF_block sampler

Description

Details of the particle filtering MCMC sampling algorithms provided in nimbleSMC.

Usage

1
2
3
sampler_RW_PF(model, mvSaved, target, control)

sampler_RW_PF_block(model, mvSaved, target, control)

Arguments

model

(uncompiled) model on which the MCMC is to be run

mvSaved

modelValues object to be used to store MCMC samples

target

node(s) on which the sampler will be used

control

named list that controls the precise behavior of the sampler, with elements specific to samplertype. The default values for control list are specified in the setup code of each sampling algorithm. Descriptions of each sampling algorithm, and the possible customizations for each sampler (using the control argument) appear below.

RW_PF sampler

The particle filter sampler allows the user to perform particle MCMC (PMCMC) (Andrieu et al., 2010), primarily for state-space or hidden Markov models of time-series data. This method uses Metropolis-Hastings samplers for top-level parameters but uses the likelihood approximation of a particle filter (sequential Monte Carlo) to integrate over latent nodes in the time-series. The RW_PF sampler uses an adaptive Metropolis-Hastings algorithm with a univariate normal proposal distribution for a scalar parameter. Note that samples of the latent states can be retained as well, but the top-level parameter being sampled must be a scalar. A bootstrap, auxiliary, or user defined particle filter can be used to integrate over latent states.

For more information about user-defined samplers within a PMCMC sampler, see the NIMBLE User Manual.

The RW_PF sampler accepts the following control list elements:

RW_PF_block sampler

The particle filter block sampler allows the user to perform particle MCMC (PMCMC) (Andrieu et al., 2010) for multiple parameters jointly, primarily for state-space or hidden Markov models of time-series data. This method uses Metropolis-Hastings block samplers for top-level parameters but uses the likelihood approximation of a particle filter (sequential Monte Carlo) to integrate over latent nodes in the time-series. The RW_PF sampler uses an adaptive Metropolis-Hastings algorithm with a multivariate normal proposal distribution. Note that samples of the latent states can be retained as well, but the top-level parameter being sampled must be a scalar. A bootstrap, auxiliary, or user defined particle filter can be used to integrate over latent states.

For more information about user-defined samplers within a PMCMC sampler, see the NIMBLE User Manual.

The RW_PF_block sampler accepts the following control list elements:


nimbleSMC documentation built on Dec. 11, 2021, 10:01 a.m.