R/glan.R

In ldbounds: Lan-DeMets Method for Group Sequential Boundaries

"glan" <- function(t,za,zb,drft){
    h <- 0.05
    stdv <- sqrt(diff(c(0,t))) # These are subroutine "sd"
    sdproc <- sqrt(t)                  # These are subroutine "sd"
    yb <- zb*sdproc-drft*t
    ya <- za*sdproc-drft*t
    nints <- ceiling((yb-ya)/(h*stdv))
    qneg1 <- pnorm(za[1],mean=drft*t[1]/stdv[1])
    qpos1 <- 1-pnorm(zb[1],mean=drft*t[1]/stdv[1])
    cp <- matrix(0,length(t),2)
    cp[1,] <- c(qpos1,qneg1)
    if (length(t) >= 2){
      grid <- seq(ya[1],yb[1],length=nints[1]+1) # These are "first"
      last <- dnorm(grid,mean=0,sd=stdv[1])      # These are "first"
      for (i in 2:length(t)){
        cpr <- cprob(last,nints,ya,yb,i,stdv[i])
        cp[i,] <- c(cpr[[1]],cpr[[2]])
        if (i < length(t)){
          hlast <- (yb[i-1]-ya[i-1])/nints[i-1]                 # These are "other"
          x <- seq(ya[i],yb[i],length=nints[i]+1)               # These are "other"
          last <- fcab(last,nints[i-1],ya[i-1],hlast,x,stdv[i]) # These are "other"
        }
      }
    }
    pr <- sum(cp)
    ans <- list(pr=pr,qpos=cp[,1],qneg=cp[,2])
    return(ans)
  }