getperform: Get performance metrics for classification

Description

Evaluate the prediction performance under the classification settings.

Usage

getperform(
  ytest,
  ypretest,
  family,
  threshold.method = c("youden", "specificity.control", "quantile"),
  specificity = NULL
)

Value

List of measures.

AUC: Area under the ROC curve.
AUPRC: Area under the precision-recall curve.
threshold: Selected threshold of the probability.
sensitivity: Sensitivity with the selected threshold.
ppv: Positive predictive value with the selected threshold.
specificity: Specificity with the selected threshold.
true.positive: Number of true positive with the selected threshold.
false.positive: Number of false positive with the selected threshold.

Arguments

ytest: Response vector for test data.
ypretest: Predicted probability for test data.
family: "ls" or "logic". Return MSE when "ls" is used.
threshold.method: Method to get the threshold.
specificity: User-defined specificity or quantile.

Details

The function supports three methods to select the threshold of the predicted probability.

threshold.method = "youden": The optimal threshold corresponds to the point that maximizes the distance to the identity (diagonal) line on the ROC curve.

threshold.method = "specificity.control": The optimal threshold corresponds to the smallest value that ensures the required specificity value.

threshold.method = "quantile": The optimal threshold corresponds to the required quantile of the predicted probability.