CI_AUC: Asymptotic confidence interval of AUC based on a variance...

In fmoyaj/aucvar: Calculate the variance of

Asymptotic confidence interval of AUC based on a variance estimator and the asymptotic normality

Description

Asymptotic confidence interval of AUC based on a variance estimator and the asymptotic normality

Usage

CI_AUC(
  formula_string,
  link = "logit",
  data,
  label_true,
  conf_level = 0.95,
  method = "unbiased",
  B = Inf,
  d = Inf
)

Arguments

formula_string	A string with an expression of the form y ~ model that represents the binary classification model. It may include operators as +, ^ and :
link	A string specifying the model link function for glm function used to fit the binomial model. Possible links are logit., probit, cauchit, (corresponding to logistic, normal and Cauchy CDFs respectively) log and cloglog (complementary log-log). The default is logit.
data	A data frame, list or environment containing the variables in the model except for the response variable. It can also be an object coercible by as.data.frame to a data frame.
label_true	A vector of the true labels in the data set, coded as 1 (positive) and 0 (negative)
conf_level	The confidence level required.The default is 0.95.
method	The method to use to compute the variance estimator. The possible methods are "unbiased" for the unbiased variance estimator of the AUC devised by Wang and Guo, "jackknife" for the jackknife variance estimator, and "bootstrap" for the bootstrap variance estimator of the AUC.
B	if the method chosen is "unbiased", B is an integer indicating the desired number of bootstrap samples to calculate the variance of the AUC. If the method chosen is "unbiased" or "jackknife" and B is set to Inf or if it is omitted, the exact number of possible partitions will be calculated.
d	Number of data entries to remove to generate the jackknife samples for the jackknife variance estimator. This argument is only required when the method chosen is "jackknife". If d is equal to 1, the delete-one version of the jackknife variance estimator will be used in which case B does not need to be specified.

Value

A matrix (or vector) with columns giving lower and upper confidence limits for each confidence level. These will be labeled as (1-level)/2 and 1 - (1-level)/2 in percentage. (2.5% and 97.5% by default).

Examples

library(aucvar)
my_data <- na.omit(breastcancer) # Omit NA values
model_formula <- "Class~`Clump Thickness`+ `Uniformity of Cell Size`+`Uniformity of Cell Shape`+
`Marginal Adhesion` + `Single Epithelial Cell Size` + `Bare Nuclei` +
`Bland Chromatin` + `Normal Nucleoli` + `Mitoses`"
# Use quotes inside double quotes since data set variable names have spaces
labels <- my_data$Class
CI_AUC(model_formula, "logit", my_data, labels, 0.95, "unbiased", B = 10^3)

fmoyaj/aucvar documentation built on Nov. 28, 2023, 10:50 p.m.