Figure 6From: Reducing decision errors in the paired comparison of the diagnostic accuracy of screening tests with Gaussian outcomes Receiver operating characteristic curves for a hypothetical oral cancer screening study. The study is subject to paired screening trial bias. The true areas under the curves for Test 1 and Test 2 are 0.77 and 0.71, respectively, for a true difference of 0.06. The observed difference is -0.06, with the corrected difference at 0.06.Back to article page