4 ms·
Ahhhh, the replication crisis. You know that isn't true in every field, right? Plus, the exact thing we're talking about HAS been replicated. Many times. Talk t
by godelski 8y ago
Ahhhh, the replication crisis. You know that isn't true in every field, right? Plus, the exact thing we're talking about HAS been replicated. Many times. Talk to a scientist about the replication scientist. Talk to ones in different fields. You'll see different responses. Correlation between the softer and harder. I'll admit particle physics has replication problems, but that's for different reasons (machines) than psychology (lots of reasons). Some fields require what would be equivalent to p values of far far less than 0.05. When we get multiple experiments agreeing when chance of coincidence is one in ten million, we tend to think that's good.
Also you're confounding the experiment. You can't do this just over New York and Montana. You do it over a lot of places. Over many countries. Plus we have other ways to test this. You can look to see if radio waves heat tissue. The fact is cell phone radiation (radio waves) are non-ionizing and unless you turn the power WAY WAY up, it isn't dangerous. Both these methods are well established and we know what we're doing. So after just these two (there's been others) we have two experiments that match our current theory.
I want to repeat that. Experiments are fitting our theory. Not hypothesis. Theory. We've been studying electromagnetism for a long time. We've been studying radio for a long time. The data is matching well established theories.
This is completely different than your jellybean case because that doesn't match any good hypothesis. There's no underlining reasoning for why only a green jellybean causes acne. That doesn't deny the jellybean correlation, but that's all it is at that point. Correlation. Which is different than causation. To graduate to causation (which is actually the thing being pointed out in that xkcd) you need to give reasoning. There are plenty of things that correlate that don't cause. [0] That's where theory comes in. But if you can't establish a good link, then you just have [0] and people laughing at you.
Tldr: go talk to scientists.
tldrr: you aren't smarter than domain experts (in their domain)
tldrrr: don't shout replication crisis when the issue we're talking about has been replicated with similar results with multile well established methods.
[0] https://tylervigen.com/spurious-correlations https://tylervigen.com/spurious-correlations