| Type: | Package |
| Title: | Simple and Fast ROC Curves |
| Version: | 1.0.3 |
| Description: | A set of functions for receiver operating characteristic (ROC) curve estimation and area under the curve (AUC) calculation. All functions are designed to work with aggregated data; nevertheless, they can also handle raw samples. In 'ROCket', we distinguish two types of ROC curve representations: 1) parametric curves - the true positive rate (TPR) and the false positive rate (FPR) are functions of a parameter (the score), 2) functions - TPR is a function of FPR. There are several ROC curve estimation methods available. An introduction to the mathematical background of the implemented methods (and much more) can be found in de Zea Bermudez, Gonçalves, Oliveira & Subtil (2014) and Cai & Pepe (2004). |
| License: | GPL-3 |
| Encoding: | UTF-8 |
| URL: | https://github.com/da-zar/ROCket |
| BugReports: | https://github.com/da-zar/ROCket/issues |
| Imports: | data.table (≥ 1.13.0) |
| Suggests: | testthat |
| RoxygenNote: | 7.2.0 |
| Collate: | 'ROCket.R' 'generics.R' 'auc.R' 'mwu.R' 'rkt_ecdf.R' 'rkt_prep.R' 'rkt_roc.R' 'roc_methods.R' 'utils.R' 'zzz.R' |
| NeedsCompilation: | no |
| Packaged: | 2025-05-09 16:06:52 UTC; daniel |
| Author: | Daniel Lazar [aut, cre] |
| Maintainer: | Daniel Lazar <da-zar@gmx.net> |
| Repository: | CRAN |
| Date/Publication: | 2025-05-09 16:20:05 UTC |
Calculate the AUC
Description
Calculate the AUC
Usage
auc(x, ...)
## S3 method for class ''function''
auc(x, ...)
## S3 method for class 'curve'
auc(x, lower, upper, n = 10000, ...)
## S3 method for class 'rkt_roc'
auc(x, exact = TRUE, ...)
Arguments
x |
An R object. |
... |
Further parameters. |
lower, upper |
The limits of integration. |
n |
The number of integration points. |
exact |
Logical. If the exact formula should be used for calculating the AUC instead of numerical approximation. |
Value
The area under the curve as a numeric value.
Mann-Whitney U test
Description
Performs the Mann-Whitney U test with a normal approximation.
Usage
mwu.test(prep, alternative = c("two.sided", "less", "greater"), correct = TRUE)
Arguments
prep |
A |
alternative |
The alternative hypothesis type. One of: "two.sided", "less", "greater". |
correct |
Logical. Whether to apply continuity correction. |
Value
A list of the class "htest".
Empirical estimate of the CDF
Description
Calculate an empirical cumulative distribution function based on a sample x and optionally a vector w of weights.
Usage
rkt_ecdf(x, w)
## S3 method for class 'rkt_ecdf'
print(x, ...)
## S3 method for class 'rkt_ecdf'
mean(x, ...)
## S3 method for class 'rkt_ecdf'
variance(x, ...)
## S3 method for class 'rkt_ecdf'
plot(x, ...)
Arguments
x |
Numeric vector containing the sample. Alternatively, if |
w |
Optional. Numeric vector containing the weights of each value in |
... |
Further parameters. |
Details
The weights vector w can contain the counts of each distinct value in x, this is the most natural use case.
In general the weights are describing the jumps of the final ecdf. Normalization is handled internally.
If x contains duplicates, corresponding values in w will be summed up.
Only positive weights are allowed. Elements in x with non-positive weights will be ignored.
Value
A function of class rkt_ecdf.
Examples
require(ROCket)
plot(rkt_ecdf(rnorm(100)))
plot(rkt_ecdf(c(0, 1)))
plot(rkt_ecdf(c(0, 1), c(1, 10)))
ROC points
Description
Calculate the ROC points for all meaningful cutoff values based on predicted scores.
Usage
rkt_prep(scores, positives, negatives = totals - positives, totals = 1)
## S3 method for class 'rkt_prep'
print(x, ...)
## S3 method for class 'rkt_prep'
plot(x, ...)
Arguments
scores |
Numeric vector containing the predicted scores. |
positives |
Numeric vector of the same length as |
negatives |
Similar to |
totals |
How many times each score was predicted. Defaults to 1 (assuming data is not aggregated).
If any value in |
x |
An environment of class |
... |
Further parameters. |
Details
In a situation where many of the predicted scores have the same value it might be easier and faster to use aggregated data.
Value
An environment of class rkt_prep.
Examples
require(ROCket)
plot(rkt_prep(1:4, c(0, 1, 0, 1)))
plot(rkt_prep(1:4, c(0, 1000, 0, 1000), totals = 1000))
plot(rkt_prep(1:4, c(100, 200, 300, 400), totals = c(1000, 800, 600, 400)))
Empirical estimate of the ROC
Description
Calculate the empirical estimate of the ROC from raw sample or aggregated data.
Usage
rkt_roc(prep, method = 1)
## S3 method for class 'rkt_roc'
print(x, ...)
## S3 method for class 'rkt_roc'
plot(x, ...)
Arguments
prep |
A |
method |
A number specifying the type of ROC estimate. Possible values can be viewed with |
x |
An object of class |
... |
Value
An object of class rkt_roc, i.e. a function or a list of two functions (for method = 1).
Examples
require(ROCket)
scores <- c(1, 2, 3, 4)
positives <- c(0, 1, 0, 1)
prep <- rkt_prep(scores, positives)
roc1 <- rkt_roc(prep, method = 1)
roc2 <- rkt_roc(prep, method = 2)
roc3 <- rkt_roc(prep, method = 3)
plot(roc1)
plot(roc2)
plot(roc3)
Available ROC estimation methods
Description
Show the implemented ROC estimation methods.
Usage
show_methods()
Value
A data.table containing the number and a short description of each implemented method.
Sample Variance
Description
Sample Variance
Usage
variance(x, ...)
## Default S3 method:
variance(x, ...)
Arguments
x |
An R object. |
... |
Further parameters. |
Value
The (biased) sample variance as a numeric value.