For regression forest, replace k folds with a single leave-one-out prediction

yixinsun1216 / crossfit

Implementation of Double/Debiased Machine Learning approach

3 stars 2 forks source link

For regression forest, replace k folds with a single leave-one-out prediction #18

Open yixinsun1216 opened 3 years ago

yixinsun1216 commented 3 years ago

Suggestion from Stephen Wager - use default regression_forest settings to get the “out of bag” predictions for each observation, skipping doing the k folds crossfitting

See Github discussion on this.
Also referenced in Athey, Tibshirani, and Wager