R/simulate_trials.R
extend.RdA trial simulation carried through a predetermined number of replications
may not achieve desired precision as judged by Monte Carlo standard errors
(MCSE) of estimated toxicity counts. This method enables simulations of
class c('precautionary','simulations') to be extended until a given
level of precision has been achieved on expected counts of enrollment and
DLTs, or (optionally) for a fixed additional number of simulations.
extend(sims, num_sims = NULL, target_mcse = 0.05)An existing object of class c('precautionary','simulations')
Optionally, a fixed number of additional replications to accumulate
Optionally, an MCSE constraint to be imposed on expected counts of DLTs, non-DLTs, and Total enrollment.
An extended simulation of same class as sims.
The MCSE constraint is imposed during trial simulation, at which
point only binary toxicities are available. Thus, as a practical matter,
extend can target MCSEs only for DLTs, non-DLTs and Total enrollment.
The subsequent subdivision of these categories during trial summary
(at which point the ordinalizer comes into play along with its parameters) thus
may generate expected counts with MCSEs exceeding target_mcse.
In practice, however, this tends to affect the estimated counts only for the
lowest toxicity grades---those of least concern from a trial-safety
perspective.
See examples under simulate_trials and format.safetytab.