STATISTICAL COMPARISON OF 2 ROC-CURVE ESTIMATES OBTAINED FROM PARTIALLY-PAIRED DATASETS

The authors propose a new generalized method for ROC-curve fitting and statistical testing that allows researchers to utilize all of the dat a collected in an experimental comparison of two diagnostic modalities , even if some patients have not been studied with both modalities. Th eir new algorithm, ROCKIT, subsumes previous algorithms as special cas es. It conducts all analyses available from previous ROC software acid provides 95% confidence intervals for all estimates. ROCKIT was teste d on more than half a million computer-simulated datasets of various s izes and configurations representing a range of population ROC curves. The algorithm successfully converged for more than 99.8% of all datas ets studied. The type I error rates of the new algorithm's statistical test for differences in A, estimates were excellent for datasets typi cally encountered in practice, but diverged from alpha for datasets ar ising from some extreme situations.