Calibrate a one-stage two-arm directional ROPE design for binary endpoints — design_twoarm_onestage_rope

Calculates equal-allocation sample sizes for a parallel-group one-stage study with binary endpoints, where the treatment effect is the risk difference pT - pC. Depending on direction, the design targets non-inferiority, superiority, or a nested procedure in which superiority is assessed only after non-inferiority has been established. Superiority is declared when the posterior probability of pT - pC > delta_sup is at least gamma_sup; non-inferiority is declared when the posterior probability of pT - pC > -delta_ni is at least gamma_ni; harmful inferiority is declared when the posterior probability of pT - pC <= -delta_ni is at least gamma_inf; otherwise the decision is inconclusive.

Usage

design_twoarm_onestage_rope_directional(
  nmin,
  nmax,
  nstep = 1L,
  direction = c("noninferiority", "superiority", "nested"),
  delta_ni,
  delta_sup = 0,
  gamma_ni = 0.975,
  gamma_sup = 0.975,
  gamma_inf = 0.975,
  analysis_prior_C = c(1, 1),
  analysis_prior_T = c(1, 1),
  design_prior_inf = NULL,
  design_prior_ni = NULL,
  design_prior_sup = NULL,
  design_prior_not_sup = NULL,
  target_ni_power = NULL,
  target_sup_power = NULL,
  target_false_ni = NULL,
  target_false_sup = NULL,
  sustainn = 1L,
  rel.tol = 1e-10,
  integration = c("adaptive", "quantile"),
  quad_nodes = 128L,
  calibration = c("Bayesian", "frequentist", "hybrid", "full"),
  freq_ni_pC = NULL,
  freq_ni_delta = NULL,
  freq_sup_pC = NULL,
  freq_sup_delta = NULL,
  target_freq_ni_power = NULL,
  target_freq_sup_power = NULL,
  target_freq_false_ni = NULL,
  target_freq_false_sup = NULL,
  freq_grid_n = 101L,
  freq_pC_range = c(0, 1),
  parallel = FALSE,
  ncores = 1L,
  parallel_backend = c("auto", "psock", "none"),
  returngrid = TRUE,
  returnmatrices = FALSE,
  progress = interactive(),
  progress_every = 1L
)

Arguments

nmin, nmax: Minimum and maximum per-arm sample size.
nstep: Positive per-arm sample-size increment.
direction: One of "noninferiority", "superiority", or "nested".
delta_ni: Non-inferiority margin on the risk-difference scale.
delta_sup: Superiority margin on the risk-difference scale, >= 0 by convention, with delta_sup > -delta_ni.
gamma_ni: Posterior threshold for non-inferiority.
gamma_sup: Posterior threshold for superiority; must be at least gamma_ni.
gamma_inf: Posterior threshold for harmful inferiority.
analysis_prior_C, analysis_prior_T: Length-two Beta analysis-prior shapes.
design_prior_inf, design_prior_ni, design_prior_sup, design_prior_not_sup: Baseline-risk/risk-difference design priors under the harmful-inferiority, non-inferiority, superiority, and not-superiority design distributions, respectively. Required depending on direction and the requested targets.
target_ni_power: Target predictive probability of a non-inferiority or superiority conclusion, evaluated under design_prior_ni.
target_sup_power: Target predictive probability of a superiority conclusion, evaluated under design_prior_sup.
target_false_ni: Maximum predictive probability of a non-inferiority or superiority conclusion, evaluated under design_prior_inf.
target_false_sup: Maximum predictive probability of a superiority conclusion, evaluated under design_prior_not_sup.
sustainn: Number of consecutive feasible evaluated sample sizes.
rel.tol: Adaptive-integration relative tolerance.
integration: Either "adaptive" or "quantile".
quad_nodes: Number of Gauss-Legendre nodes for quantile quadrature.
calibration: One of "Bayesian", "frequentist", "hybrid", or "full".
freq_ni_pC, freq_ni_delta: Control response and true difference for pointwise frequentist non-inferiority power.
freq_sup_pC, freq_sup_delta: Control response and true difference for pointwise frequentist superiority power.
target_freq_ni_power, target_freq_sup_power: Target pointwise frequentist powers.
target_freq_false_ni, target_freq_false_sup: Maximum boundary-grid frequentist error rates.
freq_grid_n: Boundary-grid size.
freq_pC_range: Control-risk interval for boundary-grid evaluation.
parallel: Logical; if TRUE, evaluate candidate per-arm sample sizes in parallel.
ncores: Positive integer giving the maximum number of worker processes used when parallel = TRUE.
parallel_backend: Parallel backend. "auto" and "psock" use a PSOCK cluster. "none" evaluates sequentially.
returngrid: Return the full operating-characteristic grid.
returnmatrices: Retain matrices for the selected design.
progress: Logical; if TRUE, show a progress bar during sequential evaluation.
progress_every: Positive integer specifying the interval, in evaluated candidate sample sizes, at which sequential progress is updated.

Value

An object of class "twoarm_rope_directional_onestage_design".