Use this tool to check whether the probabilities predicted by a model match the outcomes that actually occur. Upload a file with predicted probabilities and observed 0/1 outcomes (Excel, CSV or SPSS), or paste them, and read the calibration curve, Brier score, slope and intercept. Restyle the figure, then sign in free to download it with an APA 7 Word report.
About this tool
A calibration curve groups predictions into bins and plots the mean predicted probability against the observed event rate. Points on the diagonal indicate perfect calibration; points below it mean the model over-predicts, and points above it mean it under-predicts.
The tool reports the Brier score, the calibration intercept and slope from a logistic recalibration model and an overall test statement, and it draws the curve with confidence intervals, a reference line and optional annotations.
When to use it
Use it when validating clinical prediction models, machine-learning classifiers and risk scores, where discrimination (AUC) alone is not enough and the predicted risks themselves must be trustworthy.
What this tool reports
Calibration curve with binned observed event rates and confidence intervals.
Brier score as an overall accuracy measure.
Calibration intercept and slope.
Reference diagonal and optional on-plot statistics.
APA 7 results text and tables for the Word report.
Assumptions to check
Outcomes are coded 0/1 and predictions are probabilities between 0 and 1.
There are enough events in each bin; very small bins give wide intervals.
Predictions are evaluated on data not used to fit the model, ideally an external sample.
How to report it in APA 7
Report the sample size, the number of events, the Brier score, and the calibration slope and intercept with their intervals, then interpret them, for example “A slope below 1 indicated overfitting.” The tool drafts this paragraph and the APA 7 table.
Downloads
Analyses and figures are free to view. Create a free account or log in to download the figure as a picture (PNG or SVG) or the combined APA 7 results and figure as a Word (.docx) file. Colours, text colour, text size and line size can be changed before you download.
Frequently asked questions
Is the calibration curve generator free?
Yes. You can analyse your data and view the figure without an account. A free account is needed only to download.
What is a good calibration slope?
A slope near 1 and an intercept near 0 indicate good calibration. A slope well below 1 suggests overly extreme predictions.
How many bins should I use?
Ten bins is the usual default; use fewer bins for small samples so that each bin holds enough events.
Is my data uploaded?
No. Your file is read in your browser and never leaves your device.