lpo_auc: Compute the leave-pair-out cross-validation estimator of AUC.

In benkeser/nlpred: Estimators of Non-Linear Cross-Validated Risks Optimized for Small Samples

Compute the leave-pair-out cross-validation estimator of AUC.

Description

This estimator is computed by leaving out a pair of one case (Y = 1) and one control (Y = 0). The learner is trained on the remaining observations and predicted values are obtained for the left-out pair. The estimate is given by the proportion of left-out pairs for which the case had higher predicted risk than the control.

Usage

lpo_auc(Y, X, learner = "glm_wrapper", max_pairs = NULL, parallel = FALSE, ...)

Arguments

Y	A numeric vector of outcomes, assume to equal 0 or 1.
X	A data.frame of variables for prediction.
learner	A wrapper that implements the desired method for building a prediction algorithm. See ?glm_wrapper or read the package vignette for more information on formatting learners.
max_pairs	The maximum number of pairs to leave out.
parallel	A boolean indicating whether prediction algorithms should be trained in parallel. Default to FALSE.
...	Other options (not currently used)

Examples

# simulate data
X <- data.frame(x1 = rnorm(50))
Y <- rbinom(50, 1, plogis(X$x1))
# compute lpo_auc for logistic regression
lpo <- lpo_auc(Y = Y, X = X, learner = "glm_wrapper")

benkeser/nlpred documentation built on June 11, 2022, 3:05 a.m.