Calls out to bbi and returns a named list of class bbi_{.model_type}_summary with model outputs and diagnostics.

model_summary(.mod, .bbi_args = NULL, ..., .dry_run = FALSE)

# S3 method for bbi_nonmem_model
model_summary(.mod, .bbi_args = NULL, ..., .dry_run = FALSE)

Arguments

.mod

Model to summarize.

.bbi_args

A named list specifying arguments to pass to babylon formatted like list("nm_version" = "nm74gf_nmfe", "json" = T, "threads" = 4). See print_bbi_args() for full list of options.

...

args passed through to bbi_exec()

.dry_run

show what the command would be without actually running it

Details

NONMEM

The returned list for a NONMEM model will contain the following top-level elements:

  • absolute_model_path -- Absolute path to the model that generated this summary. Note, using this directly is discouraged in favor of using the "getters" described in the get_path_from_object help page.

  • run_details -- General details about the run including estimation method, numbers of patients and records, significant digits, run time, and more.

  • run_heuristics -- Boolean flags indicating potential issues with the run. For example, if there was a large condition number or a parameter near boundary. TRUE for a given element indicates that issue was detected. FALSE can either mean it was not detected or it was not applicable for that run.

  • parameters_data -- Data about parameter estimates. This can be accessed directly, but is much easier to look at with the param_estimates() function (which parses this, and other related elements, into a tibble).

  • parameter_names -- Names of parameters (THETA1, THETA2, etc.). Parsed into tibble by param_estimates().

  • ofv -- Objective function value, with and without constant. List with one element for each estimation method.

  • condition_number -- Condition number, if eigenvalues were returned from NONMEM. List with one element for each estimation method.

  • shrinkage_details -- Information about shrinkage. The shrinkage values (using SD parameterization) are parsed into the tibble output from param_estimates(). List of lists with one element for each estimation method, and one element per sub-population within that.

The summary call will error if it does not find certain files in the output folder. However, you can override this behavior with the following file-specific flags:

  • no_ext_file

  • no_grd_file

  • no_shk_file

If you are using an estimation method that does not produce any of the following files, or they are missing for some other legitimate reason, pass the appropriate flags through the .bbi_args argument. For example, if have asked to skip the $COV step, you would call model_summary(..., .bbi_args = list(no_cov_file = TRUE)).

Additionally, if you have renamed the .ext file from its default of <root>.ext you will need to pass ext_file = "NEWNAME" to .bbi_args.

Methods (by class)

  • bbi_nonmem_model: Get model summary from bbi_nonmem_model object

See also