PPI++ estimator for discrete predictions

Uses closed-form optimal lambda (no constraints): $$\lambda^* = \frac{\mathrm{Cov}\left(Y_{\text{cal}},\hat{Y}_{\text{cal}}\right)}{\mathrm{Var}(\hat{Y}_{\text{cal}}) + \frac{m}{n} \cdot \mathrm{Var}(\hat{Y}_{\text{test}})}$$

Usage

ppi_pp_discrete_point_and_ci(Y_cal, Yhat_cal, Yhat_test, alpha = 0.1)

Arguments

Y_cal

Vector of true labels on the calibration set.

Yhat_cal

Vector of surrogate predictions on the calibration set.

Yhat_test

Vector of surrogate predictions on the test set.

alpha

Miscoverage level for the confidence interval.