CausalForestDML to get ate and ate confidence interval on training data

[DailiZhang2010]

There are two ways to do it:

1. use the est.ate_ and est.ate_stderr_. Under the hood, it uses Doubly Robust ATE on training data
2. use est.ate(X=X, T0=T0, T1=T1) and est.ate_interval(X=X, T0=T0, T1=T1)

My target is to get the ate and confidence interval on the training data set.

The question is: which one is more reliable?

Thanks for the great package and awesome documentation.

[fverac]

I would recommend ate_ for your use case, since you should get tighter confidence intervals.

See more info here #753 (comment)

This is expected; as you note the ate_ attribute applies a double-robustness correction to the computation of the ATE itself (on the training data); the ate() method allows you to compute the ATE for any population by averaging the computed CATE values for each individual, so will not provide exactly the same result; however, if your use case is to compute the ATE for a data set that was not used in training then only the ate() method can be used for that.

[DailiZhang2010]

Thanks, @fverac