Project workbench
One dataset, the whole line: baseline table → prediction model → ROC / calibration / DCA → manuscript material. The integrated one-paste workbench is being ported to English; in the meantime each step below opens the matching tool, which performs the same computation locally in your browser.
The fully integrated workbench — where you paste data once and predicted probabilities flow automatically from modelling into ROC, calibration and DCA — is live on the Chinese site and is being translated. The English tools below already cover every step end to end.
Enter the project dataset
Paste a header row plus one record per line (tab- or comma-separated, straight from Excel). Code the outcome 0/1. Column types (numeric / 0-1 / categorical) are detected automatically and stored only in your browser.
Baseline table (Table 1)
Pick a grouping column; continuous variables are summarised as mean ± SD, categorical as n (%), with the between-group P value.
Prediction-model pipeline
Fit a Logistic (or Cox) model for the outcome; the predicted probabilities then drive ROC (discrimination), Hosmer-Lemeshow + calibration slope (calibration) and DCA (net benefit). A high AUC does not imply accurate probabilities — look at all three.
Manuscript material
Once a model is fitted, editable Statistical methods and Results drafts are written from your numbers, paired with a TRIPOD reporting checklist. Always check every number and add study-design details before submission.
Read the user guide for the full walkthrough, or see the methodology & references behind each engine. To browse every calculator, go to the Labs home.