Double machine learning (DML) is becoming an increasingly popular tool for automatic model selection in high-dimensional settings. Paul Hünermund discusses that DML is very sensitive to the inclusion of even a few “bad controls” in the covariate space.
Image courtesy of interviewee
