ROCtest: Generic function to build ROCtest

In jknowles/EWStools: Tools for automating the testing and evaluation of education early warning system models

Description

Explore the accuracy of classifiers using ROC as the metric

Getting an ROCtest on a generalized linear model

Getting an ROCtest on a train object

Getting an ROCtest on a caretEnsemble object

Getting an ROCtest on a caretEnsemble object

Usage

ROCtest(mod, testdata, ...)

## S3 method for class 'glm'
ROCtest(mod, testdata, ...)

## S3 method for class 'train'
ROCtest(mod, testdata, ...)

## S3 method for class 'caretEnsemble'
ROCtest(mod, testdata, ...)

## S3 method for class 'caretStack'
ROCtest(mod, testdata, ...)

Arguments

mod	A model object to generate an ROCit for
testdata	A dataframe to generate the ROC for the mode on
...	optional additional parameters to pass to pROC::coords such as best.weights

Details

The object has the following items

• thresh - the threshold for the ROC

• auc - the area under the curve

• confusematrix - the confusion matrix for the ROC fit

• rarepercent - percent of rare class correct

• falsepositive - percent of false rare class identifications

• modtype - the class of the model object

• modcall - the call to the model fitting function

• datatype - whether the ROC was computed on the "train" or the "test" data

Value

A ROCit object

Note

ROC coordinates are unsmoothed.

Author(s)

Jared E. Knowles

jknowles/EWStools documentation built on May 19, 2019, 11:42 a.m.