Evidence-appraisal glossary
Calibration in the Large
Calibration in the large asks a simple question: on average, does a model's predicted risk match the actual event rate in the group? If it predicts ten percent but twenty percent have events, it is miscalibrated in the large.
Also called: mean calibration.
It compares the mean predicted probability against the observed proportion of outcomes. A mismatch usually means the underlying event rate differs from where the model was built, for instance a higher-risk population or a different era of care. It is the most basic calibration check, is often the first thing to fail when a model moves to a new setting, and can frequently be corrected by updating the model's intercept.
This is a plain-language methodology definition for reading research. It is general education, not medical advice.