Risk glossary

In statistics, calibration is the process of adjusting the values of the parameters of a parametric model to ensure the model will output data that, for a given set of input data, matches as closely as possible data found empirically. A danger of using too many parameters is that the model will fit the empirical data too well, meaning the model will not accurately predict results for a different set of input data. This is known as overfitting.

Click here for articles on calibration.