run_adnuts: Run the adnuts MCMC for the model found in the directory...
In pacific-hake/hake-assessment: Stock Assessment for Pacific Hake

run_adnuts

R Documentation

Run the adnuts MCMC for the model found in the directory given

Description

Run the adnuts MCMC for the model found in the directory given

Usage

run_adnuts(
  path,
  num_chains = NULL,
  seed = 42,
  num_samples = 8000,
  num_warmup_samples = 250,
  adapt_delta = 0.95,
  run_extra_mcmc = FALSE,
  hess_step = TRUE,
  fn_exe = ss_executable,
  overwrite = TRUE,
  fn_logfile = model_output_log_fn,
  input_files = ss_input_files
)

Arguments

`path`	Directory where the model files reside
`num_chains`	The number of chains to run in parallel. If `NULL`, 1 less than the number of cores on the machine will be used 1 less than the number of cores on the machine will be used
`seed`	The random seed used to draw the random seeds for each chain
`num_samples`	The number of samples to output
`num_warmup_samples`	The warmup samples (equivalent of burnin)
`adapt_delta`	The target acceptance rate. See `adnuts::sample_admb()`
`run_extra_mcmc`	If `TRUE`, run SS extra mcmc option which outputs files into the `sso` subdirectory. If `FALSE`, those files will not be created and the `posteriors.sso` and `dervied_posteriors.sso` files will be in the running directory
`hess_step`	Logical. If `TRUE`, use the `hess_step` algorithm'
`fn_exe`	The name of the executable which was built using ADMB
`overwrite`	Logical. If `TRUE` and in an interactive session, don't ask user if they want to overwrite if the directory already exists, just do it. If non-interactive, the output will be overwritten and this will be ignored
`fn_logfile`	The filename of the logfile
`input_files`	The input files for SS

Details

path is the directory in which the MLE will be run, a subdirectory of this called mcmc is where the MCMC will be run using the NUTS algorithm. Inside the mcmc directory, several temporary subdirectories will be created, one for each MCMC chain labeled ⁠chain_*⁠, AkA CPU number used in the parallel execution. These will disappear once the run has completed and the output has been merged.