simVBIA: Simulate estimation after blinded interim analyses

In floatofmath/blindConfidence: Simulation of estimation after blinded interim analysis

Description

simVBIA simulates runs adaptive trials with blinded interim analysis and assesses the bias of the variance and mean estimates. Sample size review is performed either using an interim estimate of the pooled variance that is adjusted based on the assumed effect size or the uncorrected lumped variance.

Usage

simVBIA(delta, sigma, d, n1, alpha = 0.025, beta = 0.2, v = 1/2,
  s = zsd(n1, d, v, alpha, beta), runs, n2 = n1, n2min = 0, n2max = Inf,
  fulldata = FALSE, cf = 0)

Arguments

delta	true effect size
sigma	true standard deviation
d	assumed effect size
n1	first stage sample size in treatment group
alpha	= .025 pre-specified alpha level
beta	pre-specified beta level
v	timing of interim analysis
s	assumed standard deviation, is more or less the same as n1
runs	number of simulation runs
n2	first stage sample size in control group
n2min	minimum secondstage sample size
n2max	maximum secondstage sample size
fulldata	return full data
cf	correction term added to the sample size rule

Value

If fulldata is FALSE simVBIA returns a list with the following items, summarizing the simulation results for designs with sample size reassessment based on the adjusted interim variance estimates:

lower.prob	Coverage probability of the lower confidence bound
upper.prob	Coverage probability of the upper confidence bound
total.prob	Coverage probability of the two sided confidence interval
mean.bias	Bias of the mean estimate
variance.bias	Bias of the variance estimate
ev	Variance of the mean estimate as estimated from the adaptive trian
exv	Variance of the mean estimate of a fixed sample trial with corresponding sample size
vm	Variance of the mean estimate in the monte carlo sample
root.mse	Root mean squared error of the effect estimate
mean.m1	Mean second stage sample size
low.m1	Lower 10 percen quantile of second stage sample sizes
n10	Probability that the second stage sample size is zero

as well as results for designs with sample size reassessment based on unadjusted interim estimates:

uc.mean.bias	Bias of the mean estimate
uc.variance.bias	Bias of the variance estimate
uc.ev	Variance of the mean estimate as estimated from the adaptive trian
uc.exv	Variance of the mean estimate of a fixed sample trial with corresponding sample size
uc.vm	Variance of the mean estimate in the monte carlo sample
uc.root.mse	Root mean squared error of the effect estimate
uc.mean.m1	Mean second stage sample size
uc.low.m1	Lower 10 percen quantile of second stage sample sizes
uc.N10	Probability that the second stage sample size is zero
tn1	mean(n1)

Examples

simVBIA(.5,1,1,n1=10,n2min=2,runs=1000)

floatofmath/blindConfidence documentation built on May 16, 2019, 1:20 p.m.