Correct one-sided trialDesignGroupSequential object — correct_one_sided

Currently in rpact::getDesignGroupSequential(), when the test is one-sided, there is no way to specify the direction of the null and alternative hypotheses. This is an open issue in rpact.. This correction is only necessary for using rpact and impart together or when exporting the correct boundaries from a one-sided trialDesignGroupSequential object when the alternative is less than the null value.

Usage

correct_one_sided_gsd(trial_design, higher_better = TRUE)

Arguments

trial_design: A trialDesignGroupSequential which was called with the argument sided = 1 (i.e. one-sided alternative).
higher_better: Logical scalar: Should the alternative be theta > theta_0 (TRUE) or theta < theta_0 (FALSE) or

Value

A trialDesignGroupSequential object with the efficacy and futility boundaries corrected and a logical scalar directionUpper indicating whether the alternative is greater than the null value (TRUE) or less than the null value (FALSE).

Examples

# One sided test: Higher values = Better
trial_design_one_sided_higher <-
  rpact::getDesignGroupSequential(
    alpha = 0.05,
    beta = 1 - 0.80,
    sided = 1,
    informationRates = c(0.50, 0.75, 1),
    typeOfDesign = "asOF",
    typeBetaSpending = "bsOF",
    bindingFutility = FALSE
  )

# Bounds assumed are for $H_{A}: \theta > \theta_{0}$
trial_design_one_sided_higher$criticalValues
#> [1] 2.537988 2.015923 1.720132
trial_design_one_sided_higher$futilityBounds
#> [1] 0.4121106 1.1345811

# One sided test: Lower values = Better
trial_design_one_sided_lower <-
  correct_one_sided_gsd(
    trial_design = trial_design_one_sided_higher,
    higher_better = FALSE
  )

# Bounds for $H_{A}: \theta < \theta_{0}$
trial_design_one_sided_lower$criticalValues
#> [1] -2.537988 -2.015923 -1.720132
trial_design_one_sided_lower$futilityBounds
#> [1] -0.4121106 -1.1345811