Effective sample sizes of a posterior summarised against a normal reference

Both effective sample sizes reported per replicate compare the posterior against a normal reference of known scale. The moment version uses the posterior variance directly; the precision version uses the variance a normal distribution would need in order to have the same 95% credible interval width. Both are expressed relative to the target study, by subtracting its per-arm sample size.

Models whose posterior summary is available exactly, or already computed, evaluate these definitions directly rather than fitting a mixture to samples drawn from the posterior.

Usage

normal_reference_ess(
  reference_scale,
  posterior_sd,
  lower,
  upper,
  sample_size_per_arm
)

Arguments

reference_scale

Reference scale, i.e. the sampling standard deviation of the target study.

posterior_sd

Standard deviation of the posterior treatment effect.

lower

Lower bound of the 95% credible interval.

upper

Upper bound of the 95% credible interval.

sample_size_per_arm

Per-arm sample size of the target study.

Value

A list with the moment and precision effective sample sizes.