Design a single-stage multi-arm clinical trial for a normally distributed primary outcome

Source: R/des_ss_norm.R

des_ss_norm() determines single-stage multi-arm clinical trial designs assuming the primary outcome variable is normally distributed. It supports a variety of multiple comparison corrections, along with the determination of A-, D-, and E-optimal allocation ratios. In all instances, des_ss_norm() computes the required sample size in each arm, and returns information on key operating characteristics.

des_ss_norm(
  K = 2,
  alpha = 0.025,
  beta = 0.1,
  delta1 = 0.5,
  delta0 = 0,
  sigma = rep(1, K + 1),
  ratio = rep(1, K),
  correction = "dunnett",
  power = "marginal",
  integer = FALSE,
  summary = FALSE
)

Arguments

K

A numeric indicating the chosen value for K, the number of experimental treatment arms. Defaults to 2.
alpha

A numeric indicating the chosen value for α, the significance level. Defaults to 0.025.
beta

A numeric indicating the chosen value for β, used in the definition of the desired power. Defaults to 0.1.
delta1

A numeric indicating the chosen value for δ1, the 'interesting' treatment effect. Defaults to 0.5.
delta0

A numeric indicating the chosen value for δ0, the 'uninteresting' treatment effect. Defaults to 0.
sigma

A numeric vector indicating the chosen value for σ, the vector of standard deviations of the responses in each arm. Defaults to rep(1, K + 1).
ratio

Either a numeric vector or a character string indicating the chosen value for r, the vector of allocation ratios to the experimental arms. Can be specified explicitly as a numeric vector, or can be specified as "A", "D", or "E" to instruct des_ss_norm() to compute the A-, D-, or E-optimal value for r. Defaults to rep(1, K).
correction

A character string indicating the chosen multiple comparison correction. Can be any of "benjamini_hochberg", "benjamini_yekutieli", "bonferroni", "dunnett", "hochberg", "holm_bonferroni", "holm_sidak", "none", "sidak", and "step_down_dunnett". Defaults to "dunnett".
power

A character string indicating the chosen type of power to design the trial for. Can be any of "conjunctive", "disjunctive", and "marginal". Defaults to "marginal".
integer

A logical variable indicating whether the computed values in n, the vector of sample sizes required in each arm, should be forced to be whole numbers. Defaults to FALSE.
summary

A logical variable indicating whether a summary of the function's progress should be printed to the console. Defaults to FALSE.

Value

A list, with additional class "multiarm_des_ss_norm", containing the following elements:

  • A tibble in the slot $opchar summarising the operating characteristics of the identified design.

  • A numeric in the slot $N specifying N, the trial's total required sample size.

  • A numeric vector in the slot $n specifying n, the vector of sample sizes required in each arm.

  • A numeric in the slot $gamma specifying the critical threshold for p-values, γ, below which null hypotheses would be rejected. Will be NA if correction is not a single-step testing procedure.

  • A numeric vector in the slot $gammaO specifying the critical thresholds for ordered p-values, γ, to use with the chosen step-wise testing procedure. Will be NA if correction is not a step-wise testing procedure.

  • A matrix in the slot $CovZ specifying the covariance matrix, Cov(Z), of the standardised test statistics.

  • Each of the input variables.

See also

build_ss_norm, gui, opchar_ss_norm, plot.multiarm_des_ss_norm, sim_ss_norm.

Examples

# The design for the default parameters
des        <- des_ss_norm()
# An A-optimal design
des_A      <- des_ss_norm(ratio = "A")
# Using the root-K allocation rule, modifying the desired type of power, and
# choosing an alternative multiple comparison correction
des_root_K <- des_ss_norm(ratio      = rep(1/sqrt(2), 2),
                          correction = "holm_bonferroni",
                          power      = "disjunctive")