As I argued in my original comment, self-reported data is unreliable for answering this question. I simply do not trust people’s ability to attribute the causal impact of diets on their health.
It seems to me like even if this isn’t relevant to our beliefs about underlying reality (“teachers think criticism works because they’re not correcting for regression to the mean, and so we can’t rely on teacher’s beliefs”) it should be relevant to our beliefs about individual decision-making (“when people desist from vegan diets, 20% of the time it’s because they think their health is worse”), which should direct our views on advocacy (if we want fewer people to desist from vegan diets (and possibly then dissuade others from trying them), we should try to make sure they don’t think their health is worse on them).
I’m honestly not sure why think that self-reported data is more reliable than proper scientific studies like RTCs when trying to shed light on this question.
Often there’s question-substitution going on. If the proper scientific studies are measuring easily quantifiable things like blood pressure and what people care about more / make decisions based off of is the difficult-to-quantify “how much pep is in my step”, then the improper survey may point more directly at the thing that’s relevant, even if it does so less precisely.
It seems to me like even if this isn’t relevant to our beliefs about underlying reality (“teachers think criticism works because they’re not correcting for regression to the mean, and so we can’t rely on teacher’s beliefs”) it should be relevant to our beliefs about individual decision-making (“when people desist from vegan diets, 20% of the time it’s because they think their health is worse”), which should direct our views on advocacy (if we want fewer people to desist from vegan diets (and possibly then dissuade others from trying them), we should try to make sure they don’t think their health is worse on them).
Often there’s question-substitution going on. If the proper scientific studies are measuring easily quantifiable things like blood pressure and what people care about more / make decisions based off of is the difficult-to-quantify “how much pep is in my step”, then the improper survey may point more directly at the thing that’s relevant, even if it does so less precisely.