ActiveMethod corrected Sept 2026 · preprint correction in progress · checked 13 Nov 2026

PCR-2

What happens to a relationship between two measures when you take the error in both seriously?

A regression method for when both of the things you measured are noisy, corrected in public view.

Why it matters
Imperfect measurements can make a relationship look stronger or weaker than it is, and a decision built on that relationship inherits the error.
What it does, and how
Errors-in-variables regression: a line fitted with allowance for error in the predictor as well as the outcome, and with a separate term for real variation between cases that is not error at all.
Where it is up to
A numerical review in September 2026 found faults in the original method and in the comparison it was tested against, and I withdrew three claims. The corrected method recovers an established statistical model; it is not a new estimator. The corrected manuscript and code are not public and have had no independent statistical review. The April 2026 preprint used the original method, and a rerun in November 2026 found its headline result needs correcting, so read it as provisional until a corrected version is posted.

Where you come in

If you are a statistician or methodologist willing to review the corrected manuscript, that review is what this project most needs.

Write to me about PCR-2