R/ncc_estimation.R

In mdbrown/survMarkerTwoPhase: Evaluate the prognostic accuracy of a marker under two phase designs

NCC_EstandSE_SP <- function(
  data, 
  cutpoint,
  predict.time, 
  nmatch,
  B0 = 500)
{  
  
  #  browser()
  
  # N = nrow(data)
  #data$vi = 1; data$wi = 1
  
  #decompose data into its three parts
  mydata.ncc = data[[1]];
  Vij.IND = data[[2]]; 
  Iik0.mat = data[[3]]

  junk=NCC.Cox.explicit.PTB.s.y(data = mydata.ncc,
                                V.IND = Vij.IND,
                                Iik0 = Iik0.mat,
                                wgtk.ptb = NULL,
                                B0 = B0,
                                Zmatch.ind = NULL, 
                                t0 = predict.time, 
                                m.match = nmatch);
  
  Ptb.result = matrix(0,B0,5)
  
  for (i in 1:B0) {
    Ptb.result[i,] <- unlist(Ptb.ROC.FUN(junk$ck,
                                         junk$CondSck.ptb[,i],
                                         junk$Fck.ptb[,i],
                                         uu0 = cutpoint,
                                         type = "cutoff", 
                                         junk$yk))
  }
  
  
  est <- c(junk$beta.est, colMeans(Ptb.result, na.rm = TRUE))
  se  <- c(junk$beta.sd,apply(Ptb.result,2,sd))
  
  est <- data.frame(t(est)); names(est) =c("coef", "AUC", "FPR","TPR","NPV","PPV")
  se <- data.frame(t(se)); names(se) =c("coef", "AUC", "FPR","TPR", "NPV","PPV")
  
  list(estimates = est,se =se, fit = junk$fit) 
  
}



NCC_EstandSE_NP <- function(
  data, 
  cutpoint,
  predict.time, 
  nmatch, 
  B0)
{  
  

  # N = nrow(data)
  #data$vi = 1; data$wi = 1
  
  #decompose data into its three parts
  mydata.ncc = data[[1]];
  Vij.IND = data[[2]]; 
  Iik0.mat = data[[3]]

  junk=NCC.NP.kn.PTB.s.y(data = mydata.ncc,
                         V.IND = Vij.IND,
                         Iik0 = Iik0.mat,
                         wgtk.ptb = NULL,
                         B0 = B0,
                         Zmatch.ind = NULL, 
                         t0 = predict.time,
                         m.match=nmatch);
  
  Ptb.result = matrix(0,B0,5)

  for (i in 1:B0) {
    Ptb.result[i,] <- unlist(Ptb.ROC.FUN(junk$ck,
                                         junk$CondSck.ptb[,i],
                                         junk$Fck.ptb[,i],
                                         uu0 = cutpoint,
                                         type = "cutoff", 
                                         junk$yk))
  }
  
  
  est <- c(junk$beta.est, colMeans(Ptb.result, na.rm = TRUE))
  se  <- c(junk$beta.sd,apply(Ptb.result,2,sd))
  
  est <- data.frame(t(est)); names(est) =c("AUC", "FPR","TPR","NPV","PPV")
  se <- data.frame(t(se)); names(se) =c("AUC", "FPR","TPR", "NPV","PPV")
  
  list(estimates = est,se =se) 
  
}