Contrast that, though, with the extremely sad state of nutrition, where an absolute difference of 0.25% gets trumpeted immediately to the media at "200% more li

Contrast that, though, with the extremely sad state of nutrition, where an absolute difference of 0.25% gets trumpeted immediately to the media at "200% more likely", even though the effects are measured through observational studies with no proper controls. A field where who you are matters much more than the data you're presenting (and it seems reasonably accepted to throw out data that disagrees with your pet hypothesis). I recognize that, in many ways, physics has it easy: there are few sciences that reduce as well as it does and that offer ways to isolate variables like it does (even if it requires multi-billion dollar machines and years to make that happen). But it is still a shining exemplar of gradual evidence building theories to find more evidence, and there are many, many fields that aren't anywhere near that.