R/logit.5Confuse.R

In lessR: Less Code, More Results

.logit5Confuse <-
function(lm.out, out, n.vars, nm, prob_cut, x.cut) {

  # confusion matrix
  tbl <- table(out[,n.vars], out[,n.vars+1], useNA="always")

  hit0 <- tbl[1,1]; hit1 <- tbl[2,2]
  mis0 <- tbl[1,2]; mis1 <- tbl[2,1]
  per0 <- hit0 / (hit0 + mis0); per1 <- hit1 / (hit1 + mis1)
  hitT <- hit0 + hit1
  tot0 <- tbl[1,1] + tbl[1,2]
  tot1 <- tbl[2,1] + tbl[2,2]
  totG <- tot0 + tot1
  perT <- hitT / totG
  per0G <- tot0 / totG
  per1G <- tot1 / totG
  ln <- nchar(nm[1])

  cat("Probability threshold for predicting ",
      levels(lm.out$model[,nm[1]])[2], ":", " ", prob_cut, "\n", sep="")
  if (!is.null(x.cut))
    cat("Corresponding cutoff threshold for ", nm[2], ": ", round(x.cut, 3),
        sep="",  "\n")
  cat("\n")
  cat(.fmtc(" ",ln+8), "Baseline         Predicted", "\n")
  .dash(51)
  cat(.fmtc(" ",ln+7), "Total  %Tot        0      1  %Correct", "\n")
  .dash(51)
  cat(.fmtc(" ",ln), "  1  ", .fmti(tot1,6), .fmt(100*per1G,1,5), " ",
      .fmti(mis1,6), .fmti(hit1,6), "   ",
      .fmt(100*per1,1), "\n")
  cat(.fmtc(nm[1],ln), "  0  ", .fmti(tot0,6), .fmt(100*per0G,1,5), " ",
      .fmti(hit0,6), .fmti(mis0,6), "   ",
      .fmt(100*per0,1), "\n")
  .dash(51)
  cat(.fmtc(" ",ln), "Total", .fmti(totG,6), .fmtc(" ",25),
      .fmt(100*perT,1), "\n")
  cat("\n")

  hit1 <- hitT - hit0
  mis1 <- tot1 - hit1
  accuracy <- ((hit0 + hit1) / (hit0 + hit1 + mis0 + mis1)) * 100
  recall <- ((hit1) / (hit1 + mis1)) * 100
  precision <- ((hit1) / (hit1 + mis0)) * 100
  cat("Accuracy:", .fmt(accuracy,2), "\n")
  cat("Sensitivity:", .fmt(recall,2), "\n")
  cat("Precision:", .fmt(precision,2), "\n")
  cat("\n")

}