nsROC-package: nsROC

Description

Tools for estimating Receiver Operating Characteristic (ROC) curves, building confidence bands, comparing several curves both for dependent and independent data, estimating the cumulative-dynamic ROC curve in presence of censored data, and performing meta-analysis studies, among others.

Arguments

Abbreviations

The following abbreviations are frequently used in this package:

ROC: Receiver Operating Characteristic
AUC: Area Under the (ROC) Curve
Sp: Specificity
Se: Sensitivity
TPR: True-Positive Rate
FPR: False-Positive Rate

Functions

`gROC`	ROC curve estimate (generalization included)
`ROCbands`	Confidence bands for ROC curves
`compareROCdep`	Comparison of k paired ROC curves
`compareROCindep`	Comparison of k independent ROC curves
`cdROC`	Cumulative/dynamic ROC curve estimate
`metaROC`	Non-parametric ROC curve estimate for meta-analysis
`plot`	Plot an ROC curve
`plot`	Plot confidence bands for an ROC curve
`plot`	Plot a time-dependent ROC curve
`print`	Print a `groc` object
`print`	Print a `rocbands` object
`print`	Print a `cdroc` object

Dataset

This package comes with a dataset of 9 papers (meta-analysis) with the number of TP (true positive), FP (false positive), TN (true negative) and FN (false negative) about the use of the Interleukin6 (IL6) as a marker for the early detection of neonatal sepsis: interleukin6.

Installing and using

To install this package:

    install.packages("nsROC")

To load the package:

    library(nsROC)

Details

The basic function of the nsROC package is the gROC function. It will estimate an ROC curve under one of these considerations: larger values of the marker are associated with a higher probability of being positive (right-sided), the opposite (left-sided) or when both smaller and larger values of the marker are associated with having more probability of being positive (both).

Confidence bands for an ROC curve estimate resulting of the previous function can be computed and displayed by the ROCbands function. Three different methods are provided to compute them.

Several paired or unpaired ROC curves can be compared with the compareROCdep or compareROCindep function, respectively. In order to compare ROC curves different statistics can be used, and to approximate the distribution of the statistic in the paired case both permutation and bootstrap procedures are computed.

Time-dependent ROC curves can be estimated by the cumulative/dynamic approach using the cdROC function. In order to deal with the right censored problem three different statistics can be considered.

Meta-analysis of ROC curves following a non-parametric approach can be performed with the metaROC function. Both the fixed-effects and random-effects model can be considered.