IIRC there was also a paper analyzing how often results in some NLP conference held up when a different random seed or hyperparameters were used. It was quite depressing.
In topics where there is less reliance on relatively small numbers of cases (as is typical for medicine), there is also less reliance on marginal, but statistically "significant", findings.
So areas such as biochemistry, chemistry, even some animal studies, are less susceptible to over-interpretation or massaging of data.
Life sciences, social sciences, eco-sciences, humanities. Engineering is mostly OK or at least not mostly false.
When you get into the humanities a lot of papers "aren't even wrong", as in, the authors don't hold themselves to any standards of logic or rigor to begin with, or aren't even making any kind of identifiable claim about the world, and don't consider rebuttals based on such criticisms to be valid.
A lot of people overlook ecology/climatology because millenarians have made it into such a live wire, but those fields generally have worse problems than medicine. For instance, in medicine going back and retroactively changing patient records for your clinical trial is taboo. It happens remarkably often, but, everyone accepts that it's not supposed to. In climatology they retroactively change climate datasets all the time and if anyone calls them out on it they just attack the critics. They don't even accept the principle that their models should predict data as measured using a constant methodology. It's meaningless in such a field to say "the hypothesis is consistent with the data" because the historical data you analyzed might be replaced with a new version that's fundamentally different.
Reminds me of college grade inflation, where STEM is pretty much unaffected by it. In most engineering programs you have to work hard for a 3.00, and be brilliant or not sleep for a 4.00. My first engineering exam had an average of 50 (open book).
