This is a publisher's correction to a previously published article. The correction addresses a factual error in the original text and figure captions related to the definition of positive and negative classes in the context of AUROC calculation.
TL;DR
This document is a publisher's correction to a previously published article in Nature Machine Intelligence. It clarifies a misstatement regarding the definition of positive and negative classes in the context of calculating the Area Under the Receiver Operating Characteristic Curve (AUROC), ensuring accurate interpretation of the study's findings.
What Problem Does This Paper Address?
This document is a correction notice, not a research paper. It addresses an error in a previously published article titled "The curious case of the test set AUROC." The error concerned the definition of positive and negative classes when calculating AUROC. Specifically, the original text incorrectly stated which class was considered positive and which was negative, potentially leading to misinterpretation of the results presented in the article.
Methods
Not applicable. This is a correction notice, not a research paper detailing methods.
Datasets
Not applicable. This is a correction notice, not a research paper detailing datasets.
Key Insights and Ideas
- Clarification of Class Definitions: The primary insight is the correction of a misstatement regarding the definition of positive and negative classes. This is crucial for the accurate interpretation of AUROC, a common metric for evaluating binary classification models.
- Importance of Precise Language: The correction highlights the critical need for precision in scientific writing, especially when defining terms and metrics that underpin experimental results. Even seemingly minor errors in definitions can lead to significant misunderstandings.
- Correction Mechanism: The notice demonstrates the process by which scientific publishers correct errors in published work, ensuring the integrity and accuracy of the scientific record.
Key Results
The key "result" of this document is the correction of the following:
- Textual Correction: The sentence in the "Single-dataset discrepancy scores" section, which previously read "...with y=1 are the negative class," has been corrected to read "Samples that have y = 1 will be termed the positive class and those with y=0 are the negative class."
- Figure Caption Correction: The descriptions of panels a and b in the caption to Fig. 2, as well as callouts within the text referring to these panels, were swapped and have now been corrected.
These corrections have been implemented in the HTML and PDF versions of the article.
Limitations and Future Work
Not discussed in the paper. This is a correction notice.
Summary updated