Run model in NONMEM — run_nlme • pharmr.extra

Run the model directly using nmfe (not through pharmpy). This is a more reliable way of running NONMEM, and it is now possible to stream stdout and stderr to file or to console, which is useful for inspection of intermediate model fit.

Usage

run_nlme(
  model,
  data = NULL,
  tables = NULL,
  full_tables = FALSE,
  id,
  path = getwd(),
  method = c("nmfe", "pharmpy", "psn"),
  nmfe = get_nmfe_location(),
  force = NULL,
  console = FALSE,
  save_fit = TRUE,
  save_summary = TRUE,
  estimation_method = NULL,
  sir_options = NULL,
  auto_stack_encounters = TRUE,
  clean = TRUE,
  as_job = FALSE,
  save_final = TRUE,
  check_only = FALSE,
  remove_tables = FALSE,
  verbose = TRUE
)

Arguments

model: pharmpy model object or NONMEM model code (character) or path to NONMEM model file.
data: filename of dataset or data.frame as input to NONMEM / nlmixr. Optional, can also be included in model object (if specified as pharmpy model object).
tables: acharacter vector of which default tables to add, options are fit and parameters. Default is NULL, i.e. don't add any new tables (but will keep existing).
full_tables: For the default tables, should all input columns from be included in the output tables? Default FALSE.
id: run id, e.g. run1. This will be the folder in which the NONMEM model is run. If no folder is specified, it will create a folder run1 in the current working directory, and will increment the run number for each subsequent run.
path: path to nonmem model. If not specified, will assume current working directory.
method: run method, either pharmpy dispatch, nmfe or psn (psn::execute).
nmfe: full path to nmfe file to run NONMEM with, if method=="nmfe".
force: if run folder (id) exists, should existing results be removed before rerunning NONMEM? Default FALSE.
console: show stderr and stdout in R console? If FALSE, will stream to files stdout and stderr in fit folder.
save_fit: save fit object. If TRUE, will save as <run_id.rds>. Can also specify filename (rds) to save to.
save_summary: save fit summary and parameter estimates to file? Default is TRUE. Will use current folder, and save as fit_summary_<id>.txt and fit_parameters_<id>.csv.
estimation_method: Optional. Character vector of estimation method(s) to apply to model. Will remove all existing estimation steps in the model and update with methods specified in argument.
sir_options: options for running SIR in covariance step. A list with options niter (number of SIR iterations) and samples (number of samples). Default NULL leaves the model unchanged. samples should be between 300 and 10000 (suggested to use 1000 by default). niter should be 1 or higher (suggest to use 1 by default).
auto_stack_encounters: only invoked if data argument supplied as a data.frame, not if a pharmpy model object is supplied without data or when data is a filename. Detects if TIME within an individual is decreasing from one record to another, which NONMEM cannot handle. If this happens, it will add a reset event (EVID=3) at that time, and increase the TIME for subsequent events so that NONMEM does not throw an error. It will increase the time for the next encounter to the maximum encounter length across all subjects in the dataset (rounded up to 100). If no decreasing TIME is detected, nothing will be done (most common case). This feature is useful e.g. for crossover trials when data on the same individual ispresent but is included in the dataset as time-after-dose and not actual time since first overall dose.
clean: clean up run folder after NONMEM execution?
as_job: run as RStudio job?
save_final: after running the model, should a file final.mod be created with the final estimates from the run.
check_only: if TRUE, will only check the model code (NM-TRAN in the case of NONMEM), but not run the model. Will return TRUE if model syntax is correct, and FALSE if not. Will also attach stdout as message attribute.
remove_tables: if TRUE, removes all $TABLE records from the model before running. Applied after any tables added via the tables argument. Default is FALSE.
verbose: verbose output?

Value

TODO

Details

The function does take a pharmpy model as input (optionally), and uses pharmpy to read the results from the model fit, and returns a pharmpy modelfit object.