Skip to contents

Thresholds at which a test's actual type I error equals given levels

Source: R/analysis_operating_characteristics.R
calibrated_levels.Rd

For each level alpha, the p-value threshold c such that the share of the null p-values strictly below c is floor(alpha * N) / N, within 1 / N of alpha. A test rejecting when its p-value is below c then has the actual rejection rate alpha in the null scenario, whatever its nominal level there. Null p-values tied at the threshold all stay unrejected, so the rate never exceeds alpha.

Usage

calibrated_levels(alpha, null_p_values)

Arguments

alpha

Levels, in [0, 1].

null_p_values

The test's p-values on the null scenario's trials.

Value

One threshold per level; Inf rejects every trial.