How to reach us
02
Write about
- A wrong number. Top priority. Send the share link and name the metric.
- A bug. Say which browser and OS.
- A missing metric or CI method.
- Citation or licensing. MIT. The Cite tab has APA and BibTeX.
03
Not this
We can't interpret your results, advise on clinical or financial decisions, or debug your pipeline. Results are informational only (disclaimer).
Before you write
Will my data be seen if I send a share link?
Yes. The link carries your counts in the URL, so anyone reading the email sees them. Nothing else leaves your browser. If the data is sensitive, use made-up numbers.
My numbers disagree with scikit-learn. Is that a bug?
Check the ordering. We put TP top left. sklearn.metrics.confusion_matrix returns [[TN, FP], [FN, TP]]. Read one as the other and sensitivity and specificity swap. Same ordering and still different? Write.
Why is my confidence interval different?
Several methods are valid. We use Wilson for proportions, log-normal (Simel/Altman) for LR+, LR− and DOR, and a normal approximation for linear combinations. Each card names its method.