It is not clear to me what are the conditions on the data used to train meta-learners.
The text says: This time, we will use non-random data to train the models and random data to validate them. Dealing with-non random data is a much harder task, because the meta learners will need to debias the data AND estimate the CATE.
If I understand correctly, we do not need to ensure that causal inference assumptions such as covariate balance on the training are valid, it is only at inference time, if we want to make sure that what we are computing is indeed the average treatment effect, that we will need to make sure the data is representative of the population the treatment would be applied to.
Is that correct ?
Hi,
It is not clear to me what are the conditions on the data used to train meta-learners. The text says:
This time, we will use non-random data to train the models and random data to validate them. Dealing with-non random data is a much harder task, because the meta learners will need to debias the data AND estimate the CATE.If I understand correctly, we do not need to ensure that causal inference assumptions such as covariate balance on the training are valid, it is only at inference time, if we want to make sure that what we are computing is indeed the average treatment effect, that we will need to make sure the data is representative of the population the treatment would be applied to. Is that correct ?Thanks