program sample(model:String <- "Model", method:String <- "ParticleFilter", input_file:String?, output_file:String?, diagnostic_file:String?, nsamples:Integer <- 1, ncheckpoints:Integer <- 1, seed:Integer?, verbose:Boolean <- true, nparticles:Integer <- 1, ess_trigger:Real <- 0.7)

Sample from a model.

General options.

  • --model: Name of the model class to use.

  • --method: Name of the method class to use.

  • --input-file: Name of the input file, if any.

  • --output-file: Name of the output file, if any.

  • --diagnostic-file: Name of the diagnostics file, if any.

  • --nsamples: Number of samples to draw.

  • --ncheckpoints: Number of checkpoints for which to run. The interpretation of this is model-dependent, e.g. for a Markov model it is the number of states.

  • --seed: Random number seed. If not provided, random entropy is used.

  • --verbose: Enable verbose reporting?

Particle filter-specific options:

  • --nparticles: Number of particles to use.

  • --ess-trigger: Threshold for resampling. Resampling is performed whenever the effective sample size, as a proportion of --nparticles, drops below this threshold.