diff_cvAUC: diff_cvAUC

In benkeser/cvAUC.plus: Estimates and Tests of Cross-validated Area Under the ROC Curve

Description

This function estimates the difference between two cross-validated AUCs fit using the wrap_cvAUC function. It also provides an influence function-based confidence interval estimate and hypothesis test of the null hypothesis that the two CV-AUCs are equal.

Usage

diff_cvAUC(fit1, fit2, confidence = 0.95)

Arguments

fit1	An object of class wrap_AUC
fit2	An object of class wrap_AUC
confidence	A numeric between 0 and 1 indicating the nominal coverage probability for the confidence interval.

Value

An object of class diff_cvAUC with the following entries:

diff	The difference in CV-AUC between the two fits.
ci	The confidence interval for the difference between the two fits.
p	The two-sided p-value for the test that the two CV-AUCs are equal.
folds	The number of folds used by fit1 and fit2.
learner1	The name of the learner used for fit1.
learner2	The name of the learner used for fit2.
confidence	The confidence interval level.

Examples

n <- 1000
X <- data.frame(x1=rnorm(n),x2=rnorm(n))
Y <- rbinom(n,1,plogis(X$x1 + X$x2))
myglm1 <- function(Y,X,newX){
   fm <- glm(Y~.,data=X,family=binomial())
   pred <- predict(fm,newdata=newX,type="response")
   return(list(fit = fm, pred = pred))
}
myglm2 <- function(Y,X,newX){
  fm <- glm(Y~x1,data=X,family=binomial())
  pred <- predict(fm,newdata=newX,type="response")
  return(list(fit = fm, pred = pred))
}
out1 <- wrap_cvAUC(Y = Y, X=X, learner = "myglm1")
out2 <- wrap_cvAUC(Y = Y, X=X, learner = "myglm2")

diff_auc <- diff_cvAUC(fit1 = out1, fit2 = out2)

benkeser/cvAUC.plus documentation built on Feb. 1, 2021, 8:42 a.m.