oob_score :bool, default=False

[sunsunsun5114]

What happens if you set oob_score :bool, default=False to True in causal forest, and how many folds will the data become?

[jbbqqf]

Hi! Triaging older issues — looks like this is a usage question rather than a bug or feature request, so it might be a candidate for closure with a quick pointer.

oob_score=True on the causal forest behaves like sklearn's RandomForestRegressor: each tree is trained on a bootstrap sample and the out-of-bag observations (the ~37% of rows not sampled) are used to compute a generalization estimate. There is no k-fold split; the data is partitioned per-tree by the bootstrap sampling. See causalml/inference/tree/causal/causalforest.py:154 for the docstring and :405-414 for where it's consumed (if self.oob_score: self._set_oob_score_and_attributes(X, y)).

For broader usage discussion the README points to GitHub Discussions / the Slack channel — that's a better fit than the issue tracker for "what does parameter X do."

(Disclosure: I drafted this comment with help from Claude Code while triaging stale issues; the references above were verified manually against the current master branch.)