Refit specified datasets in a stansim_simulation object

refit Takes a stansim_simulation object and a vector of characters corresponding to the names of datasets fitted within the stansim_simulation object, and refits the stan model for each of these instances. This allows users to refit any specific models using new stan arguments if need be (e.g. if the model fails to converge in the original run).

refit(object, datasets, stan_args = list(), calc_loo = FALSE,
  use_cores = 1L, cache = TRUE, seed = floor(stats::runif(1, 1, 1e+05)))

Arguments

object	An object of S3 class stansim_simulation.
datasets	The full names of the data files to be refitted. These must be consistent both with the dataset names stored within the stansim_simulation object, and with the copies of the data files relative to the current working directory. This is best ensured by running refit from the same working directory as the original fit_models() call.
stan_args	A list of function arguments to be used by the internal rstan::sampling() function when fitting the models. If not specified then the sampling() defaults are used.
calc_loo	If TRUE then model fit statistics will be calculated using the loo package. If TRUE there must be a valid log_lik quantity specified in the generated quantities section of the provided stan model.
use_cores	Number of cores to use when running in parallel. Each stan model is fitted serially regardless of the number of chains ran as parallelisation across models is more flexible than within.
cache	If TRUE then the results for each instance are written to a local, temporary file so that data is not lost should the function not terminate properly. This temporary data is removed upon the model terminating as expected. If FALSE no data is written and results are only returned upon the correct termination of the whole function. The default value of TRUE is recommended unless there are relevant write-permission restrictions.
seed	Set a seed for the fit_models() function.

Value

An S3 object of class stansim_simulation recording relevant simulation data.

Examples

# NOT RUN {
# refit datasets "data_file-12.rds" & "data_file-08.rds"
refit(simulation,
      datasets = c("data_file-12.rds", "data_file-08.rds")
      use_cores = 4)

# refit dataset "data_file-12.rds" using a larger number of samples
refit(simulation,
      datasets = "data_file-12.rds",
      stan_args = list(iter = 4000),
      use_cores = 4)
# }