Calculate the Brier Score — calc

Calculates the Brier Score, i.e. MSE of probabilities in [0, 1] of a vector of probabilities. The brier score attempts to account for not only wither a prediction correctly predicts a class label at some arbitrary evaluation cutoff, but also how close the prediction is to predicting the label, i.e. distinguishing between \(p = 0.51\) and \(p = 0.99\), despite both predicting a positive class label.

Usage

calc_brier(x, p)

Arguments

x: numeric(n). A vector of binary class data representing the true classes. Must be all 0 or 1 or numeric coercible.
p: numeric(n). A vector of the predicted probabilities, i.e. in [0, 1].

Value

The Brier Score, a value in [0, 1], representing the error in predictions, 0 being best possible score.

References

https://en.wikipedia.org/wiki/Brier_score

Author

Stu Field

Examples

withr::with_seed(1, {
  n <- 100L
  p <- runif(n)
  x <- sample(0:1, n, replace = TRUE)
})
calc_brier(x, p)
#> [1] 0.3005455