Comment by strombofulous | Hacker News Reader

Comment by strombofulous | Hacker News Reader

I am really suspicious of this 99.9% number. Most journals allow a ~5% margin of error so even if humans caused climate change (which they most likely did) we should only see 95% of these studies finding this result. Also this doesn't account for the fact that people and organizations have their own biases and that it's possible to construct "experiments" to give you whatever result you want, so we should be seeing some experiments showing that climate change isn't real or whatever, bringing that 95% number down even further.

If this number is accurate and not just chosen for PR reasons, it would lend credence to the idea that anti-human-climate-change research is being suppressed. Is this bad? Probably not, but it's probably a more accurate conclusion than declaring climate change the correct conclusion.

> Most journals allow a ~5% margin of error so even if humans caused climate change (which they most likely did) we should only see 95% of these studies finding this result.

Well, for a phenomenon strong enough, any measurement you make will find it. It's like taking a 1m long board and trying to assert the hypothesis that it's 20cm long. It doesn't matter that you accept a 0.05 p-value, you don't have an instrument noisy enough to reach it.

Did you even read the article? It gives a brief explanation of the methodology. The full paper is found here: https://iopscience.iop.org/article/10.1088/1748-9326/ac2966

I'm not saying that the authors were dishonest I'm saying that if this is an honest outcome from the paper's authors (which it probably is) then the result we should expect should be closer to 95% than 99.9%. That it's this off from the expected value indicates that something external to this paper is inaccurate, even if these authors measured the inaccuracies correctly.

That's not how statistical significance works. :) By the same logic, 5% of all historical studies on the Holocaust should state that it is a hoax or else Holocaust denial is being suppressed. Furthermore 5% is a very low significance level, way below what is used by most reputable journals.

So if a study were done on whether humans could fly (without planes/gliders), there'd be 5% of studies finding that people can indeed fly, and publish that as real science?

Reply on news.ycombinator.com