Repository navigation
Confounding adjustment language in docstrings #149
Description
Activity
- added a commit that references this issue
on Oct 6, 2026 Fixed in #150. The adjusted estimators' docstrings and examples (which now simulate randomized treatment), together with the Oregon tutorial, now describe ML adjustment as variance reduction in randomized experiments, not as confounding correction. The scope is stated explicitly in the docstrings.
Generated by Claude Code
This issue is in relation to JOSS review thread openjournals/joss-reviews#11047
I notice some of the docstrings discuss ML models being used for confounding adjustment (AdjustedDistributionEstimator class in simple.py, for example, including also the given example which simulates confounding in the treatment). While this is of course a reasonable application under selection on observables, the underlying theory seems to be all about variance reduction in the case of treatments/instruments orthogonal to the additional covariates. I suggest to either clarify the scope in the docstring or make the confounding adjustment use case more prominent across the package and simulations.