eval: Evaluate the Performance of the Prediction

In SkadiEye/ITRlearn: Individualized Treatment Rule Learning

Description

Evaluate the performance of the prediction using different criteria.

Usage

evalPred(object, test.Y, cutoff = 0.5, type = c("Binary", "Continuous")[1],
  output = ifelse(type == "Continuous", list("mse"), list(c("acc", "sens",
  "spec", "auc", "ppv", "npv", "w.acc", "a.acc", "val.f")))[[1]],
  weight = rep(1, length(test.Y)), inclusion = rep(TRUE, length(test.Y)),
  Y.val = test.Y, Y.resp = rep(1, length(test.Y)))

## S4 method for signature 'PredObj'
evalPred(object, test.Y, cutoff = 0.5,
  type = c("Binary", "Continuous")[1], output = ifelse(type == "Continuous",
  list("mse"), list(c("acc", "sens", "spec", "auc", "ppv", "npv", "w.acc",
  "a.acc", "val.f")))[[1]], weight = rep(1, length(test.Y)),
  inclusion = rep(TRUE, length(test.Y)), Y.val = test.Y, Y.resp = rep(1,
  length(test.Y)))

## S4 method for signature 'ListPredObj'
evalPred(object, test.Y, cutoff, type, output, weight,
  inclusion, Y.val)

Arguments

object	A PredObj or a ListPredObj object.
test.Y	The true response.
cutoff	A cutoff for binary response.
type	The type of the response, either "Binary" or "Continuous".
output	The output types. See details.
weight	Sample weight for weighted accuracy.
inclusion	Used in thresholded accuracy.
Y.val	Used in the value function.

Details

acc: Accuracy sens: sensitivity spec: specificity auc: area under the curve ppv: positive predictive value npv: negative predictive value w.acc: weighted accuracy — w.acc = sum((Y==pred.Y)*weight)/sum(weight) a.acc: thresholded accuracy — a.acc = sum((Y==pred.Y)*inclusion)/sum(inclusion) val.f: value function — val.f = sum((Y.val==pred.Y)*weight)/sum(weight)

Value

A data.frame of performance will be retured.

See Also

dwnnel-predict

SkadiEye/ITRlearn documentation built on May 24, 2019, 1:31 a.m.