filter_traj: Filtering trajectories

In kidusasfaw/pomp: Statistical Inference for Partially Observed Markov Processes

Description

Trajectories drawn from the smoothing distribution

Usage

## S4 method for signature 'pfilterd_pomp'
filter.traj(object, vars, ...)

## S4 method for signature 'pfilterList'
filter.traj(object, vars, ...)

## S4 method for signature 'pmcmcd_pomp'
filter.traj(object, vars, ...)

## S4 method for signature 'pmcmcList'
filter.traj(object, vars, ...)

Arguments

object	result of a filtering computation
vars	optional character; names of variables
...	ignored

Details

The smoothing distribution is the distribution of

Xt | Y1=y1*, …, YT=yT*,

where Xt is the latent state process, Yt is the observable process, t is time, and T is the time of the final observation.

In a particle filter, the trajectories of the individual particles are not independent of one another, since they share ancestry. However, a randomly sampled particle trajectory X_1,…,X_T is a draw from the smoothing distribution. Seting filter.traj = TRUE in pfilter causes one such trajectory to be sampled. By running multiple independent pfilter operations, one can thus build up a picture of the smoothing distribution.

In particle MCMC (pmcmc), this operation is performed at each MCMC iteration. Assuming the MCMC chain has converged, and after proper measures are taken to assure approximate independence of samples, filter.traj allows one to extract a sample from the smoothing distribution.

See Also

Other particle filter methods: bsmc2, cond.logLik, eff.sample.size, filter.mean, mif2, pfilter, pmcmc, pred.mean, pred.var

kidusasfaw/pomp documentation built on May 20, 2019, 2:59 p.m.