When should an author retract their paper as a result of misinterpreted results?

Question

I'm curious when should an author make a call on retracting their paper vs. sending an errata or a correction. Many say it is when an error is involved that influences the interpretation of the paper.

I'm finding it hard to understand when a paper should or shouldn't be retracted, since there are plenty of papers out there with errors at every step of the process but they are still available. There are also several papers that have been outright wrong but they are still available. There are several papers that have been performed incorrectly, interpreted incorrectly, and still published in a high profile journal that haven't been retracted.

This is distinct from retractions due to fraud or other ethical issues.

Answer 1

Outside of ethical issues, I'd argue that the only time that retraction of a paper is warranted, is when there is a methodological error that is so great that it completely invalidates all or almost all of the paper's key results. For example, if a paper is supposed to observe phenomena A, B, C, and D, I wouldn't retract the paper if the experiment that was supposed to measure D was done incorrectly, but the results of A, B, and C remain unchanged. A correction or erratum should suffice in such cases. However, if a paper was trying to derive X, Y, and Z, but faulty assumption Q invalidates all of the derived results, then that paper should be retracted.