Recover the number of Monte Carlo replicates behind a binomial estimate

Operating characteristics are stored as summaries (estimate, Monte Carlo standard error, exact confidence interval) rather than as the underlying counts. Propagating their uncertainty correctly requires the replicate count, which this function recovers from whichever summary is informative.

The Monte Carlo standard error identifies the replicate count exactly whenever it is available and non-zero. It is zero precisely when no replicate succeeded or all of them did, and in that case one bound of the exact interval is a closed-form function of the replicate count alone. As a last resort, when no standard error was stored for an interior estimate, the count is approximated from the width of the interval.

Usage

binomial_replicate_count(
  estimate,
  mcse,
  conf_int_lower,
  conf_int_upper,
  conf_level = 0.95
)

Arguments

estimate

The estimated proportion.

mcse

The Monte Carlo standard error of the estimate, or NA.

conf_int_lower

Lower bound of the exact (Clopper-Pearson) interval.

conf_int_upper

Upper bound of the exact (Clopper-Pearson) interval.

conf_level

The confidence level the interval was computed at.

Value

The number of replicates, or NA_real_ when no summary identifies it.