ARmethod {BayesianPlatformDesignTimeTrend}

ARmethod

Description

This function adjusts the posterior randomisation probability for each arm using many approaches. Currently Thall's approach and Trippa's approach are used. Double biased coin and other method will be added in the next version.

Usage

ARmethod(
  BARmethod,
  group,
  stats,
  post.prob.btcontrol,
  K,
  n,
  tuningparameter = NA,
  c = NA,
  a = NA,
  b = NA,
  post.prob.best,
  max.ar = NA,
  armleft,
  treatmentindex
)

Arguments

BARmethod

The indicator of which adaptive randomisation method is used

group

The current stage

stats

The output matrix

post.prob.btcontrol

The vector of posterior probability of each active treatment arm better than control

K

Total number of arms at the beginning

n

The vector of sample size for each arm

tuningparameter

The tuning parameter indicator for Thall's approach

c

The tuning parameter for Thall's approach

a

The hyperparamter parameter for Trippa's approach

b

The hyperparamter parameter for Trippa's approach

post.prob.best

Posterior probability of each arm to be the best

max.ar

The upper boundary for randomisation ratio for each arm, which is used in Thall's approach since Trippa's approach has protection on control arm.

armleft

The number of treatment left in the platform (>2)

treatmentindex

The vector of treatment arm index excluding the control arm whose index is 0

Value

randomprob: The vector of adjusted randomisation probability to each arm

Author(s)

Ziyan Wang

References

Bayesian adaptive randomized trial design for patients with recurrent glioblastoma. Trippa, Lorenzo, Eudocia Q. Lee, Patrick Y. Wen, Tracy T. Batchelor, Timothy Cloughesy, Giovanni Parmigiani, and Brian M. Alexander. Journal of Clinical Oncology 30, no. 26 (2012): 3258. A simulation study of outcome adaptive randomization in multi-arm clinical trials. Wathen, J. Kyle, and Peter F. Thall. Clinical Trials 14, no. 5 (2017): 432-440.

Examples

ARmethod(
BARmethod = "Thall",
group = 1,
stats = matrix(rep(NA, 40), ncol = 8, nrow = 5),
post.prob.btcontrol = 0.5,
K = 2,
n = c(30, 30),
tuningparameter = "fixed",
c = 1,
post.prob.best = c(0.5, 0.5),
max.ar = 0.75,
armleft = 2,
treatmentindex = 1)

ARmethod(
BARmethod = "Trippa",
group = 1,
stats = matrix(rep(NA, 40), ncol = 8, nrow = 5),
post.prob.btcontrol = c(0.5, 0.6),
K = 3,
n = c(30, 30, 40),
tuningparameter = NA,
c = NA,
a = 3,
b = 0.75,
post.prob.best = c(0.3, 0.3, 0.4),
max.ar = NA,
armleft = 3,
treatmentindex = c(1, 2))