This answer just seems plain wrong to me.
The question states that there are more features than observations, so backwards stepwise selection (which would require fitting the full model, which would not converge) just doesn't seem doable.
However, two considerations.
- The author is a diamond mod, so as a junior member of CV I would err on the side of being deferential, and someone already pointed out the problem. The author's response didn't make a lot of sense to me. ("This is machine learning, so you can.")
- If the question were not a CW I think I would just -1 it, +1 the existing objection, and move on. In this case though, I am wondering whether the traditional injunction against editing someone else's answer in a way that is inconsistent with the intent still holds here. (Related question, but not on CW answers: Should I edit a factual error?)