proc.ci: Partial AUC Inference

Description

Infer the area of region under ROC curve with pre-specific FPR constraint (FPR-pAUC). See http://www3.stat.sinica.edu.tw/statistica/j27n1/j27n117/j27n117.html for details.

Usage

proc.ci(response, predictor, cp = 0.95, threshold = 0.9, method = "MW")

Arguments

response

a factor, numeric or character vector of responses; typically encoded with 0 (negative) and 1 (positive). Only two classes can be used in a ROC curve. If its levels are not 0/1, the first level will be defaultly regarded as negative.

predictor

a numeric vector of the same length than response, containing the predicted value of each observation. An ordered factor is coerced to a numeric.

numeric; coverage probability of confidence interval.

threshold

numeric; false positive rate (FPR) constraint.

method

methods to estimate FPR-pAUC. MW: Mann-Whitney statistic. expect: method in (2.2) http://www.ncbi.nlm.nih.gov/pubmed/20729218. jackknife: jackknife method in http://www3.stat.sinica.edu.tw/statistica/j27n1/j27n117/j27n117.html.

Value

Confidence interval of FPR partial AUC.

Details

This function infers FPR partial AUC given response, predictor and pre-specific FPR constraint. MW: Mann-Whitney statistic. method in http://www3.stat.sinica.edu.tw/statistica/j27n1/j27n117/j27n117.html adapted from http://www.ncbi.nlm.nih.gov/pubmed/20729218. jackknife: jackknife method in http://www3.stat.sinica.edu.tw/statistica/j27n1/j27n117/j27n117.html.

Examples

Run this code


library('pROC')
data(aSAH)
proc.ci(aSAH$outcome, aSAH$s100b, cp=0.95 ,threshold=0.9,method='expect')