pmcmc: Run a Particle-MCMC
In StateSpaceModels/ssminr: ssminr

Function to run a MCMC using a SMC (particle filter) to evaluate the log-likelihood.

pmcmc(
  ssm,
  approx = c("ode", "sde", "psr"),
  dt = NULL,
  id = 0,
  root = NULL,
  iter = NULL,
  n_parts = NULL,
  n_thread = 1,
  n_obs = NULL,
  cooling = NULL,
  switch = NULL,
  eps_switch = NULL,
  eps_max = NULL,
  eps_abs_integ = NULL,
  eps_rel_integ = NULL,
  smooth = FALSE,
  alpha = NULL,
  like_min = NULL,
  freeze_forcing = NULL,
  interpolator = NULL,
  verbose = FALSE,
  warning = FALSE,
  no_dem_sto = FALSE,
  no_white_noise = FALSE,
  no_diff = FALSE,
  traj = TRUE,
  n_traj = NULL,
  hat = FALSE,
  trace = TRUE,
  acc = TRUE,
  seed_time = TRUE
)

`ssm`	a `ssm` object, returned by `new_ssm`.
`approx`	character, approximation used to simulate `ssm`: `"ode"` deterministic approximation, based on Ordinary Differential Equations. `"sde"` diffusion approximation, based on Stochastic Differential Equations. `"psr"` Euler-multinomial approximation, based on Poisson process with Stochastic Rates.
`dt`	numeric, integration time step.
`id`	integer, unique integer identifier that will be appended to the output.
`root`	character, root path for output files (if any) (no trailing slash). If `NULL` (default), outputs are written in "your_model_path/the_name_of_the_wrapper".
`iter`	numeric, number of iterations.
`n_parts`	numeric, number of particles.
`n_thread`	numeric, number of threads to be used. Default to 1. Use `"max"` to set it automatically to the number of cores available on the machine using `detectCores`.
`n_obs`	numeric, number of observations to be fitted (for tempering). If `NULL` (default), all observations are fitted.
`cooling`	numeric, cooling factor (for sampling covariance live tuning or `mif` cooling).
`switch`	numeric, select switching iteration from initial covariance to empirical one (`kmcmc` and `pmcmc`) or to update formula introduced in Ionides et al. 2006 (`mif`)
`eps_switch`	numeric, select number of burn-in iterations before tuning epsilon.
`eps_max`	numeric, maximum value allowed for epsilon.
`eps_abs_integ`	numeric, absolute error for adaptive step-size control.
`eps_rel_integ`	numeric, relative error for adaptive step-size control.
`smooth`	logical, tune epsilon with the value of the acceptance rate obtained with exponential smoothing. Default to `FALSE`.
`alpha`	numeric, smoothing factor of exponential smoothing used to compute smoothed acceptance rate (low values increase degree of smoothing)
`like_min`	numeric, particles with likelihood smaller than `like_min` are considered lost. If `NULL` (default) lower bound on likelihood based on machine precision.
`freeze_forcing`	character, freeze covariates to their value at specified date (in `YYYY-MM-DD` format).
`interpolator`	character, gsl interpolator for covariates
`verbose`	logical, print logs (verbose). Default to `FALSE`.
`warning`	logical, print warnings. Default to `FALSE`.
`no_dem_sto`	logical, turn off demographic stochasticity (if any). Default to `FALSE`.
`no_white_noise`	logical, turn off white noises (if any). Default to `FALSE`.
`no_diff`	logical, turn off diffusions (if any). Default to `FALSE`.
`traj`	logical, print the trajectories. Default to `TRUE`.
`n_traj`	numeric, number of trajectories stored. If `NULL` (default), one trajectory is stored at each iteration.
`hat`	logical, print the state estimates. Default to `FALSE`.
`trace`	logical, print the trace. Default to `TRUE`.
`acc`	logical, print the acceptance rate. Default to `TRUE`.
`seed_time`	logical, seed the random number generator with the current time. Default to `TRUE`.