bootstrappedMetricCurve: bootstrappedMetricCurve

In Peder2911/evallib: Simple binary classifier evaluation

Description

Makes metricCurve data using bootstrapping, which makes it possible to calculate the standard deviation, and the quantiles of the results.

Usage

bootstrappedMetricCurve(pred, actual, x, y, res = 0.1, draws = 100,
  parallel = FALSE)

Arguments

pred	Probability predictions
actual	Outcomes
res	Resolution of ROC thresholds. A float between 0 and 1.
draws	Number of times to randomly sample the data for bootstrapping. Larger numbers yield better results, but also increase runtime.
parallel	Boolean. Run in parallel? Uses ncores - 1 CPU cores.
probs	Vector of quantiles to return

Details

Note that the function requires specification of the resolution of the metricCurve. This is because each bootstrap iteration must have the same dimensions, since it must be possible to collapse these into an array.

Value

A list containing ROC data and AUC with confidence intervals.

Examples

tdat <- examplePredictions()
bootstrappedMetricCurve(tdat$pred,tdat$actual,roc)

Peder2911/evallib documentation built on Dec. 18, 2019, 2:41 a.m.