R/getnorm.pool.R

In TULIP: A Toolbox for Linear Discriminant Analysis with Penalties

getnorm.pool <- function(x, y){
  f0 <- ecdf(x[y==1])
  f1 <- ecdf(x[y==2])
  n0 <- sum(y==1)
  n1 <- sum(y==2)
  n <- n0+n1
  delta.n0 <- 1/n0^2
  delta.n1 <- 1/n1^2
  
  v0 <- rep(0,n)
  v0[y==1] <- qnorm(truncate(x[y==1], f0, delta.n0))
  v0[y==2] <- qnorm(truncate(x[y==2], f0, delta.n0))
  mu0.hat <- mean(v0[y==2])
  
  v1 <- rep(0,n)
  v1[y==1] <- qnorm(truncate(x[y==1], f1, delta.n1))
  v1[y==2] <- qnorm(truncate(x[y==2], f1, delta.n1))
  mu1.hat <- mean(v1[y==1])
  mu.hat <- n0/n*mu0.hat-n1/n*mu1.hat  
  
  transform <- function(t){
    n0*qnorm(truncate(t,f0,delta.n0))/n+n1*(qnorm(truncate(t,f1,delta.n1))+mu.hat)/n
  }
  x.norm <- transform(x)
  
  list(x.norm=x.norm, f0=f0, f1=f1, mu.hat=mu.hat, transform=transform)
}