R/score_direct.R

In MatheMax/OptReSample: Computing Optimal Adaptive Designs

#' Compute the score
#'
#' \code{score_direct} computes the objective score approximately if the \code{c_2}-function and the \code{n_2}-function
#' are approximated via a set of splines.
#'
#' Note that nodes, c2 and n2 have to be of the same length.
#' The score is given as the expected sample size of the design under the alternative hypothesis.
#' The integrals are approximated via the trapezial rule.
#'
#'
#' @param parameters Parameters (alpha, power, standardized effect) whith which you want to build your design
#' @param cf Boundary for stopping for futility after the first stage
#' @param ce Boundary for stopping for efficacy after the first stage
#' @param nodes Nodes for the stage two functions
#' @param c2 c_2-values that correspond to \code{nodes}
#' @param n1 First stage sample size
#' @param n2 n_2-values that correspond to \code{nodes}
#'
#' @export


score_direct <- function(parameters, cf, ce, nodes, c2, n1, n2){
  g <- splinefun(nodes, n2)

  N=12

  h = (ce - cf) / N
  x = seq(cf, ce, h)
  alpha = c(1, rep(2,(N-1)), 1)

  # w = c(x, alpha)
  sc <- function(w){
    y <- w[2] * g(w[1]) * dnorm( w[1] - sqrt(n1) * parameters$mu)
  }

  y <- apply(cbind(x,alpha), 1, sc)
  p <- (h/2) * sum(y)
  p <- p + n1
  return(p)
}