Specialty › Power (hypothesis tests)

Compare two AUCs

Power analysis for two paired AUCs uses the Hanley-McNeil correlated standard error.

What is Compare two AUCs?

For paired AUCs (same patients, two markers), Hanley-McNeil tabulated the correlation between AUCs as a function of the correlation between scores (ρ); we use the user-supplied ρ directly. Higher ρ ⇒ more power (the paired design exploits the within-patient correlation).

When should I use Compare two AUCs?

Powering a head-to-head diagnostic comparison.

What data does it need?

AUC₁ + AUC₂ + κ + ρ + α + n or power.

What does it report?

n_pos or achieved power.

How do I interpret the result?

Pair design with high score-correlation (ρ > 0.7) cuts required n dramatically vs ρ = 0 — when feasible, run both tests on the same patients.