Generate a synthetic PMX dataset from a fitted model

Draws new subjects from the fingerprint of a synpmx_model_estimate() fit. This stage reads no patient data: its arguments are the model and a subject count, so everything about the source that reaches the output has already passed through the fit. A full fit is reduced to its model_fingerprint() first, so generation cannot read a diagnostic the fingerprint leaves behind.

Usage

synpmx_model_generate(fitted_model, n_subjects = NULL, seed = NULL)

Arguments

fitted_model

A fingerprint from model_fingerprint(), or the pmx_fitted_model from synpmx_model_estimate() it was taken from.

n_subjects

Number of synthetic subjects. Defaults to the source count.

seed

Generation seed.

Value

A data frame in the source's shape, carrying the fingerprint it was generated from as its pmx_fitted_model attribute.

Details

Per subject: an arm is assigned keeping the source arm shares, covariates are drawn from the study-wide covariate model, random effects from the between-subject covariance matrix, and the dose schedule from the arm's dosing model. The concentration is then evaluated at the visits drawn from the arm's visit model, against the schedule that was drawn, so a reduced or skipped dose reaches the concentrations rather than appearing only in the dosing records. A PD-only fit skips concentration simulation: responses use their fitted study-time shapes or visit frequencies, while any declared dosing records are still generated.

See also

synpmx_model_estimate(), model_fingerprint(), synpmx_model(), model_report().