パーティション（ディシジョンツリー）で定義される交差検証の統計量はどのようなものでしょうか？

[#1]

K-fold cross-validation randomly divides the (non-excluded) rows in a data table D into k subsets D1, D2…, Dk.
The subsets are of equal size and independent of each other.

K distinct models are trained using data from D-Di, where Di is the split that was removed.

For continuous responses, the error for each observation in Di is calculated using the model trained from D-Di.
For nominal or ordinal responses, JMP calculates Gˆ2 from each observation in Di using a model trained on D-Di.
This is repeated for each of the K partitions.

The resulting errors are squared and then added (for nominal and ordinal responses, the Gˆ2 values ​​are added).
This is how the cross-validated SSE (Gˆ2 for nominal or ordinal responses) is constructed.

FAQ # 2097