4 ms·
Correct for it as best as possible (eg. by correcting for every demographic factor you can), and then for results that still look promising, do actual controlle
by londons_explore 3y ago
Correct for it as best as possible (eg. by correcting for every demographic factor you can), and then for results that still look promising, do actual controlled studies - eg. "contact mcdonalds and get them to halve the amount of additive X in food they serve in any zip code ending in an even digit".
If additive X still looks like it's doing harm when the data from the controlled study comes in, ban it.
- amanaplanacanal 3y agoThe controlled studies are expensive, and almost never actually get done. Which is why most diet related science is so bad.
- londons_explore 3y agoBut as a government, you can force the biggest users of an additive to do a controlled study by a deadline date or have authority to use the substance at all in food revoked.
- gruez 3y agoBut in this case it seems like confounders is guaranteed to turn up a massive amount of false positives. Forcing companies to fund said studies on shaky evidence seems arbitrary and unfair.