AucDiff: Calculate difference between areas under the ROC curve (AUC) between a forecast and a reference forecast for the same observation, and estimate the variance of the AUC difference

Description

Calculate difference between areas under the ROC curve (AUC) between a forecast and a reference forecast for the same observation, and estimate the variance of the AUC difference

Usage

AucDiff(
  fcst,
  fcst.ref,
  obs,
  handle.na = c("na.fail", "only.complete.triplets"),
  use_fn = c("C++", "R")
)

Arguments

fcst

vector of forecasts

fcst.ref

vector of reference forecasts

obs

vector of binary observations (0 for non-occurrence, 1 for occurrence of the event)

handle.na

how should missing values in forecasts and observations be handled; possible values are 'na.fail' and 'only.complete.triplets'; default: 'na.fail'

use_fn

the function used for the calculation: 'C++' (default) for the fast C++ implementation, or 'R' for the slow (but more readable) R implementation

Value

vector with AUC difference, and estimated standard deviation

References

DeLong et al (1988): Comparing the Areas under Two or More Correlated Receiver Operating Characteristic Curves: A Nonparametric Approach. Biometrics. https://www.jstor.org/stable/2531595 Sun and Xu (2014): Fast Implementation of DeLong's Algorithm for Comparing the Areas Under Correlated Receiver Operating Characteristic Curves. IEEE Sign Proc Let 21(11). 10.1109/LSP.2014.2337313

Examples

Run this code

# NOT RUN {
data(eurotempforecast)
AucDiff(rowMeans(ens.bin), ens.bin[, 1], obs.bin)
# }