A serialized cmlnmr() fit retains the individual patient data in its refit
arguments and observed-outcome slots. redact_fit() removes those, so a
saved or shared object carries no row-level data; the posterior draws,
component design, diagnostics, and estimability information are preserved.
Arguments
- object
A
cmlnmr()fit.
Details
After redaction the object can no longer be refitted, so prior_sensitivity()
will not run on it. Under either backend the sampler object in fit$fit may
still hold the model data it was sampled with, so for a fully data-free
artifact save only what you need from the posterior, such as the output of
posterior_summary(), rather than the fit itself.