Letter to the editorDetermining sample size for roc studies: what is reasonable for the expected difference in tests’ ROC areas?
References (5)
- et al.
Components-of-variance models and multiple-bootstrap experimentsan alternative method for random-effects, receiver operating characteristic analysis
Acad Radiol
(2000) - et al.
A method of comparing the areas under Receiver Operating Characteristic curves derived from the same cases
Radiology
(1983)
Cited by (16)
Two-dimensional frontal plane projection angle can identify subgroups of patellofemoral pain patients who demonstrate dynamic knee valgus
2018, Clinical BiomechanicsCitation Excerpt :AUC values range from 0 to 1.0, with a score of 1.0 having the ability to perfectly discriminate between conditions. Classification of AUC scores can be interpreted as excellent (0.90 to 1.0), good (0.80 to 0.90), fair (0.70 to 0.60), and weak (>0.50), with a cut-off level of 0.50 used to indicate a failed point of sensitivity (Obuchowski, 2003). Subjects in the PFP and uninjured group were well matched for age, height, weight and foot length (Table 2).
How Many Readers and Cases does One Need to Conduct an ROC Study?
2011, Academic RadiologyAn Approach to Comparing Accuracies of Two Flair MR Sequences in the Detection of Multiple Sclerosis Lesions in the Brain in the Absence of Gold Standard
2010, Academic RadiologyCitation Excerpt :We observed greater differences in JAFROC-1 FOMs than in the corresponding ROC areas under the curve, as expected, because FOMs range from 0 to 1, whereas areas under ROC curves usually vary from 0.5 to 1. Because, for large AUC values, even small differences (in the order of 0.05) is believed to represent a fairly large difference in performance (21), the large observed difference is of great clinical relevance for diagnosis as well as for disease monitoring in clinical trials, where effects of similar magnitude in MS lesion load on MRI are frequently observed when treatment effects are compared (1,4,22). Our approach differs from previous methods, which relied on likelihood of observer agreement (23), in that we estimated performance of observers on one TE sequence against the surrogate ground truth of the other TE sequence.
Prediction Accuracy of a Sample-size Estimation Method for ROC Studies
2010, Academic RadiologyNarrowing of the Middle Cerebral Artery: Artificial Intelligence Methods and Comparison of Transcranial Color Coded Duplex Sonography with Conventional TCD
2010, Ultrasound in Medicine and BiologyCitation Excerpt :However, an explanation should be provided to make more sense for clinicians what a “small” difference of 0.05 in ROC areas means in terms of a difference in a more clinically useful measure, such as sensitivity at a fixed false-positive rate. In opinion of Obuchowski (Obuchowski 2003) in most medical imaging studies, where performance is already likely to be reasonably high, an improvement in ROC area of 0.10 is unlikely and even a difference of 0.05 represents a fairly larger difference in sensitivities, 0.10 to 0.14. Such situation applies to our study as the sensitivity of the TCCS for detection of “any MCA narrowing” was higher than sensitivity of TCD with the use of all methods of data analysis.
The use of receiver operating characteristic curves in biomedical informatics
2005, Journal of Biomedical Informatics