adapt1 <- function(simulateP,nsims,postcut,hypothesisP,nCuts){
   win <- logical(nsims)
   ss <- numeric(nsims)

   nInts <- nCuts
   for (i in 2:length(nCuts)){
      nInts[i] <- nCuts[i] - nCuts[i-1]
   }

   for (i in 1:nsims){
      x <- rbinom(length(nCuts),nInts,simulateP)
      x <- c(x[1],x[1]+x[2],x[1]+x[2]+x[3])
      ProbSup <- 1 - pbeta(hypothesisP,1+x,1+nCuts-x)

# Probability of Success
      win[i] <- any(ProbSup > postcut )
# Sample size
      ss[i] <- min(nCuts[ProbSup > postcut],max(nCuts))
   }
   out <- c(length(win[win])/nsims,mean(ss),
                       sqrt(var(ss)),table(ss)/nsims)
   names(out) <- c('Pr(win)','MeanSS','SD SS',as.character(nCuts))
   out
}

## Input these values for the Type I error of Example 5.2.1:
> nsims <- 1000000
> postcut <- 0.95
> hypothesisP <- 0.5
> simulateP <- 0.5
> nCuts <- c(50,75,100)
> out <- adapt1(simulateP,nsims,postcut,hypothesisP,nCuts)

> out
  Pr(win)    MeanSS     SD SS        50        75       100
 0.095770 96.455625 12.247579  0.059165  0.023445  0.917390