Power for PPI++ two-sample test with binary outcomes (sens/spec input)

Computes PPI++ power for a two-sample difference in proportions when predictions come from a binary classifier summarized by sensitivity and specificity. This helper converts the classification metrics into the required moment inputs and delegates to power_ppi_ttest().

Usage

power_ppi_ttest_binary(
  p_A,
  p_B,
  n_A,
  n_B,
  N_A,
  N_B,
  sens_A,
  spec_A,
  sens_B = sens_A,
  spec_B = spec_A,
  delta = NULL,
  alpha = 0.05
)

Arguments

p_A, p_B

Group-specific prevalences \(P(Y=1)\).

n_A, n_B

Labeled sample sizes per group.

N_A, N_B

Unlabeled sample sizes per group.

sens_A, spec_A

Sensitivity and specificity for group A predictions.

sens_B, spec_B

Sensitivity and specificity for group B predictions (defaults match group A).

delta

Optional effect size \(p_A - p_B\). If NULL, it is computed from p_A and p_B.

alpha

Two-sided significance level (default 0.05).

Value

Scalar power in [0, 1] with a "moments" attribute containing the per-group moment lists.