Hanley-McNeil 1982 AUC confidence interval

Computes the closed-form Hanley-McNeil 1982 standard error and Wald confidence interval for a binary-classification AUC from the AUC estimate and the number of positive and negative samples. This is the single-sample scoring bank's default per-cell AUC interval helper; it is conservative for cross-validated estimates because it does not model fold dependence.

Usage

singlesample_hanley_mcneil_auc_ci(auc, n_pos, n_neg, conf_level = 0.95)

Arguments

auc

Numeric scalar in [0, 1].

n_pos

Integer count of positive/case samples.

n_neg

Integer count of negative/control samples.

conf_level

Numeric confidence level in (0, 1). Default 0.95.

Value

Named numeric vector with ci_lo, ci_hi, and se. Returns NA values when inputs are not estimable.

References

Hanley JA, McNeil BJ. (1982) The meaning and use of the area under a receiver operating characteristic (ROC) curve. Radiology 143(1): 29-36.

Examples

singlesample_hanley_mcneil_auc_ci(auc = 0.72, n_pos = 50, n_neg = 50)
#>      ci_lo      ci_hi         se 
#> 0.62007917 0.81992083 0.05098095