trueposrate: True positive classification
In townleym/mMisc: Matt Miscellaneous

Description Usage Arguments Details Examples

View source: R/mMisc.R

Returns the rate at which true positives are identified in a classification problem given a certain threshold. Especially for logistic/poisson regression whether a model correctly classifies a binary outcome depends on the threshold.

1	trueposrate(threshold, truth, fitted)

`threshold`	a value between 0 and 1 for the value at which an outcome is classified
`truth`	a vector of [0 \| 1] values representing the true classification
`fitted`	a vector in the range [0,1] values representing the predicted classification

This function is primarily used to calculate values for the Receiver Operating Characteristics (ROC) curve function.

m0 = glm(outcome ~ predictor1 + predictor2, family = binomial(), data = train)
testprobs = predict(m0, newdata = test, type = "response")
truth = train$outcome
trueposrate(0.5, truth, testprobs)
# Most useful for understanding the rate of true positives across a range of thresholds
sapply(0:10 / 10, function(x) {trueposrate(x, truth = truth, fitted = testprobs)})