normative
Abandon observational designs for tiny-signal nutrition questions
For questions where the noise is many orders of magnitude larger than the signal—such as single nutrients linked to disease outcomes—observational designs should be abandoned because they cannot recover such tiny effects; randomized or experimental trials minimizing confounding are needed, and continuing observational studies is largely a waste that may bias our thinking rather than provide preliminary insight.
normativepending
Speaker
John IoannidisEvidence Quote
“I would argue for this type of questions, where the error is 50 times bigger than the signal, we need to find designs that minimize the error.”
Created: 6/13/2026, 12:20:10 AM
Your notes
Loading notes...