How to calculate confidence intervals for average impact over a subset of data?

[EgorKraevTransferwise]

Most EconML models offer a way to generate a confidence interval for the point estimates of causal impact. What I need is to calculate confidence intervals over subsets of points, eg split by the estimated impact. Example: I want to estimate the confidence interval for the impact averaged over the data points where the model predicts positive impact. Can you suggest any way to do so, at least for some models, that is less compute-intensive than just bootstrapping? Thanks a lot!